CN108550174A - A kind of coastline Super-resolution Mapping and system based on half global optimization - Google Patents

Abstract

The present invention proposes a coastline super-resolution mapping method and system based on semi-global optimization. The method mainly includes: acquiring an initial coastline image and a reference image, performing image preprocessing, image fusion and image registration, combining the change trend of the overall coastline shape with the gray level change around the initial coastline, and extracting coastline change control points , and use it to divide the initial coastline into several coastline segments; in each segment, in the coastline neighborhood window, obtain the sub-pixel positioning results, and combine the sub-pixel positioning coordinates of all points in the same segment to fit it as A smooth curve segment; combining all the curve segments together is the complete coastline vector result, and completes the super-resolution mapping of the coastline; the invention proposes a sub-pixel positioning method based on local areas and can adapt to high curvature The coastline environment has high positioning accuracy and can adapt to coastlines with different curvatures, improving the accuracy of the results.

Description

A coastline super-resolution mapping method and system based on semi-global optimization

technical field

The invention belongs to the field of coastline image processing, and relates to a coastline super-resolution mapping method and system based on semi-global optimization.

Background technique

The coastline is the dividing line between the sea and the land, and is one of the 27 most important land surface features. Affected by natural factors and human factors, the coastline has undergone changes in shape and nature. Changes in natural conditions such as seawater dynamics (including tidal currents, etc.), geological and meteorological disasters, and sea level rise caused by climate warming lead to coastal silting and erosion; impacts of human activities such as reclamation, land reclamation, and marine engineering, The trend and length of the coastline have also changed significantly while leading to the transformation of the coastline type.

Therefore, coastline mapping and change detection have become the basic work of coastal erosion monitoring, coastal resource management, environmental protection of coastal areas and sustainable development of coastal areas. my country is rich in marine resources and has a high degree of coastline utilization. Research on the rapid and accurate extraction of coastlines is of great practical significance for the sustainable development of society, economy and nature in my country. The main methods of coastline extraction are image segmentation method and edge detection method. The image segmentation method is to divide the image data into two parts, foreground and background, and the junction of them is the extracted coastline.

There are many image segmentation algorithms, such as using level set model to extract coastline from SAR image (Cerimele et al. 2009; Ouyang, Chong, and Wu.2010; Shu, Li, and Gomes.2010); extracting coastline based on geometric active contour model ( Niedermeier, Romaneessen, and Lehner.2000; Xing, Fu, and Zhou.2012; Zhang et al. 2013); combining image segmentation methods and image local features to extract coastlines from high-resolution images (Di et al.2003; Liu et al. al.2011); integrating supervised and unsupervised classification methods for coastline extraction (Sekovski et al.2014).

Edge detection methods extract coastlines by searching for energy intensity (brightness) discontinuities between adjacent pixels (Buono et al. 2014; Fugura, Billa, and Pradhan 2011; Zhang et al. 2013b).

Other methods include extracting coastlines from high-resolution images using mathematical morphology (Puissant et al.2008); based on photogrammetry technology, using digital aerial photos and digital orthophotos to extract coastlines (Lira et al. Automatic detection of coastlines (Valentini.2017).

So far, most of the research has been to find an optimal way to determine the location of the coastline. Since most methods are based on hard classification, the coastline only has pixel-level positioning accuracy, which cannot meet the actual accuracy requirements. Some scholars use high-resolution remote sensing images to extract coastlines. Although high-precision coastlines can be obtained, the high price cannot meet the actual needs of large-scale coastline extraction. Therefore, there is a need for a method for extracting high-precision coastlines from medium and low-resolution remote sensing images. With the development of image processing technology, especially the development of sub-pixel positioning and super-resolution reconstruction technology, some scholars use it in the coastline extraction of medium and low resolution remote sensing images, so that the extracted coastline has higher accuracy.

Foody et al. (2005) evaluated a soft classification method for coastlines from degraded simulated Landsat images, and the results showed sub-pixel-scale coastlines and an RMSE of 2.25 m. Pardo-Pascual et al. (2012) proposed an automatic method to extract coastlines with sub-pixel accuracy from Landsat TM and ETM+, and the RMSE of coastline positions ranged from 4.69 to 5.47 m. As natural beaches change over time, Almonacid-Caballer et al. (2016) applied the annual average coastline position extracted from Landsat images to substantially reduce the short-term variation of the coastline, and the extracted coastline was about 4 to 5 meters away from the sea surface.

These methods are often complex and difficult to implement in practical applications, and have certain limitations, such as being affected by the suspended sediment on the shore, unable to adapt to the high-curvature coastline environment, and unable to remove the influence of moored ships.

Most methods are based on hard classification, so the coastline only has pixel-level positioning accuracy, which cannot meet the actual accuracy requirements. In the method of pixel-level coastline extraction, when the image segmentation algorithm extracts the coastline, in order to ensure the extraction accuracy, more post-processing steps are required to determine the boundary pixels and the threshold value; the region growing algorithm must consider many criteria and thresholds, including splitting , merging, and starting point selection; the drawback of edge detection-based methods is that the discontinuities produced by these methods do not represent coastlines well. Some scholars use high-resolution remote sensing images to extract coastlines. Although high-precision coastlines can be obtained, the high price cannot meet the actual needs of large-scale coastline extraction. Therefore, there is a need for a method for extracting high-precision coastlines using low- and medium-resolution remote sensing images. However, the soft classification method is often complicated and difficult to implement in practical applications, and has certain limitations, such as being affected by the suspended sediment on the shore, unable to adapt to the high curvature coastline environment, and unable to remove the influence of moored ships.

Contents of the invention

Aiming at the shortcomings and deficiencies of the existing technical problems, the present invention provides a coastline super-resolution mapping method and system based on semi-global optimization, which is used to solve the technical problems that the positioning accuracy is not high and cannot adapt to the high-curvature coastline environment. Described method mainly comprises the steps:

S1. Obtain the Landsat 8 land imager image and the GF-2 reference image, perform image preprocessing on the Landsat 8 land imager image and the GF-2 reference image, respectively, and obtain the Landsat 8 fusion image and the water index gray corresponding to the Landsat 8 fusion image Degree map and fusion image of GF-2;

S2, No. GF-2 fusion image is used as reference image, registration is carried out to described reference image and Landsat 8 fusion image, obtains the offset parameter between reference image and Landsat 8 fusion image;

S3. Perform pixel-level coastline extraction on the water body index grayscale image to obtain an initial coastline; perform coastline extraction on a reference image to obtain a reference coastline;

S4. Extracting coastline change control points from the initial coastline to obtain initial coastline change control points, and segmenting the initial coastline through the initial coastline change control points to obtain several segmented coastlines;

S5. Carry out sub-pixel positioning based on the local area for each pixel point in each section of coastline, and obtain the sub-pixel positioning coordinates of each point in each section of coastline; through the offset parameters described in step S2 for each The sub-pixel positioning coordinates of the point are offset corrected;

S6. Carry out coastline least squares fitting to the offset-corrected sub-pixel positioning coordinates of all points in the same coastline, and fit it into a smooth curve segment; combine all curve segments together to obtain a complete Coastline vector results, complete super-resolution mapping of coastlines.

In a coastline super-resolution mapping method based on semi-global optimization of the present invention, it also includes: respectively calculating and analyzing the position errors between the coastline sub-pixel positioning coordinates and the reference coastline, between the coastline vector result and the reference coastline error result.

In a kind of coastline super-resolution mapping method based on semi-global optimization of the present invention, the sub-pixel positioning based on local area described in step S5 comprises the following steps:

S51. Calculate the edge direction of each pixel point in each section of the initial coastline, and determine different neighborhood windows for each pixel point of the initial coastline according to different edge directions;

S52. Describe the domain window boundary as a curve through the fitting function, calculate the parameters of the fitting function, and determine the coordinates of the sub-pixel positioning points in the local area.

In a kind of coastline super-resolution mapping method based on semi-global optimization of the present invention, coastline least squares fitting described in step S6 comprises the following steps:

S61. Detect and extract the sub-pixel positioning coordinates of each section of coastline;

S62. Divide the sub-pixel positioning coordinates into different point sets, and perform least squares fitting on the sub-pixel positioning coordinates in the same point set to obtain sub-pixel-level coastlines.

Preferably, the present invention also provides a coastline super-resolution mapping system based on semi-global optimization, including the following modules:

The image preprocessing image fusion module is used to obtain Landsat 8 land imager images and GF-2 reference images, respectively perform image preprocessing on Landsat 8 land imager images and GF-2 reference images, and obtain Landsat 8 fusion images, Landsat 8 The grayscale image of the water body index corresponding to the fusion image and the fusion image of GF-2;

The image registration module is used to use the GF-2 fusion image as a reference image, register the reference image and the Landsat8 fusion image, and obtain the offset parameters between the reference image and the Landsat 8 fusion image;

The coastline extraction module is used to extract the pixel-level coastline from the water body index grayscale image to obtain the initial coastline; to carry out the coastline extraction to the reference image to obtain the reference coastline;

The coastline change control point extraction is used to extract the coastline change control point for the initial coastline to obtain the initial coastline change control point, and the initial coastline is segmented by the initial coastline change control point to obtain several segmented coastlines;

The sub-pixel positioning module is used to perform sub-pixel positioning based on the local area for each pixel point in each section of coastline, and obtain the sub-pixel positioning coordinates of each point in each section of coastline; through the image registration module The above-mentioned offset parameters are used to correct the offset of the sub-pixel positioning coordinates of each point;

The positioning point fitting module is used to perform coastline least squares fitting on the offset-corrected sub-pixel positioning coordinates of all points in the same section of coastline, and fit it into a smooth curve segment; Combined to get a complete coastline vector result, complete the super-resolution mapping of the coastline.

In the coastline super-resolution mapping system based on semi-global optimization of the present invention, an error analysis module is also included: for calculating the distance between the coastline sub-pixel positioning coordinates and the reference coastline, and the distance between the coastline vector result and the reference coastline. position error and analyze the error results.

In a coastline super-resolution mapping system based on semi-global optimization of the present invention, the sub-pixel positioning based on the local area described in the sub-pixel positioning module includes the following modules:

The neighborhood window determination module is used to calculate the edge direction of each pixel point in each section of the initial coastline, and determine the different neighborhood windows of each pixel point of the initial coastline according to different edge directions;

The sub-pixel anchor point coordinate acquisition module is used to describe the domain window boundary as a curve through the fitting function, calculate the parameters of the fitting function, and determine the coordinates of the sub-pixel anchor point in the local area.

In a kind of coastline super-resolution mapping system based on semi-global optimization of the present invention, the coastline least squares fitting described in the positioning point fitting module includes the following modules:

Extract sub-pixel positioning coordinates module, which is used to detect and extract sub-pixel positioning coordinates on each coastline;

The sub-pixel level coastline acquisition module is used to divide the sub-pixel positioning coordinates into different point sets, and carry out least squares fitting to the sub-pixel positioning coordinates in the same point set to obtain the sub-pixel level coastline.

The present invention proposes a coastline super-resolution mapping method and system based on semi-global optimization, which acquires initial coastline images and reference images, performs image preprocessing, image fusion and image registration, and integrates the change trend of the overall coastline form with the initial coastline Combined with the surrounding gray level changes, the coastline change control points are extracted, and the initial coastline is divided into several coastline segments; in each segment, the sub-pixel positioning results are obtained in the coastline neighborhood window, and combined with the same segment The sub-pixel positioning coordinates of all points are fitted into a smooth curve segment; all curve segments are combined together to be a complete coastline vector result, and the super-resolution mapping of the coastline is completed; the present invention proposes a method based on The sub-pixel positioning method in the local area can adapt to the coastline environment with high curvature, and has high positioning accuracy, and can adapt to coastlines with different curvatures, which improves the accuracy of the results.

Description of drawings

The present invention will be further described below in conjunction with accompanying drawing and embodiment, in the accompanying drawing:

Fig. 1 is the flow chart of the method of the embodiment of the present invention;

Fig. 2 is the image map of the experimental area 1 of the embodiment of the present invention;

Fig. 3 is the image map of experimental area 2 of the embodiment of the present invention;

Fig. 4 is the initial coastline extraction figure of the embodiment of the present invention;

Fig. 5 is a schematic diagram of extracting coastline inflection points according to an embodiment of the present invention;

Fig. 6 is a sub-pixel positioning result diagram according to an embodiment of the present invention;

Fig. 7 is the subsection fitting diagram of the embodiment of the present invention;

Fig. 8 is the result figure of experimental zone 1 of the embodiment of the present invention;

Fig. 9 is a graph showing the results of the experimental area 2 of the embodiment of the present invention.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present invention clearer, the present invention will be described in further detail below in conjunction with the accompanying drawings and examples.

The present invention proposes a coastline super-resolution mapping method and system based on semi-global optimization. The complete flow chart of the method is shown in Figure 1. The research area of the embodiment of the present invention is divided into two parts: Caofeidian Port and Xiamen-Quanzhou surrounding coastal areas area. Caofeidian Port is located near 118.5°E, 39°N, adjacent to China's Beijing-Tianjin-Hebei urban agglomeration, and is one of China's important ore transportation ports. Caofeidian was just a sand island in the past, but it became a port after artificial construction. Therefore, its main coastline type is artificial coastline, and the coastline position has been stable for a long time and has not been changed for many years. The Xiamen-Quanzhou surrounding experimental area is located between 118°E-118.5°E and 24.35°N-24.6°N, adjacent to the Taiwan Strait, and Xiamen is an important port for international economic and cultural exchanges. The increasingly close connection with the world has promoted the coastal development around Xiamen and Quanzhou, resulting in rapid changes in the status of the coastline in recent decades. The coastline types around Xiamen and Quanzhou mainly include bedrock coast, artificial coast and flat sandy coast.

In the research of the embodiment of the invention, all experimental images are downloaded from the USGS database, and the images are acquired by the Landsat8 land imager sensor. The Landsat-8 land imager images use the WGS84 ellipsoid model and the Transverse Mercator (UTM) projection. The detailed parameters of the Landsat-8 land imager images are shown in Table 1. Table 1 summarizes the error statistics (mean error, standard deviation) for each date, each experimental region, and each data type. The mean error is obtained by averaging all the errors, and since all errors are obtained by calculating the absolute value of the distance from the final shoreline point to the reference shoreline, the mean error is used to account for the level of deviation from the reference shoreline. The standard deviation (STDEV) represents the variability around the mean error.

Table 1 Landsat-8 land imager parameter list

Caofeidian Port is located in block 122-033. We processed 12 images of this area from 2013 to 2016, and selected three artificial coastlines with different water and soil distributions from each image as the experimental area 1. It is used to determine the band combination method with the best noise immunity, as shown in Figure 2. The coastal area around Xiamen-Quanzhou was acquired from the block 119-043, and the image was acquired at 2:33 am on October 13, 2015 (UTC). Five coastlines with different curvatures were selected from the image, as shown in the figure 3, used to verify the universality of the algorithm.

In addition, the Gaofen-2 optical remote sensing satellite (GF-2) image was used to extract the reference coastline covering the study area, as shown in the figure. The Gaofen-2 satellite was successfully launched on August 19, 2014. It is China's first civilian optical remote sensing satellite with a spatial resolution better than 1 meter. It is equipped with two high-resolution 1-meter panchromatic and 4-meter multispectral cameras. It has the characteristics of sub-meter spatial resolution, high positioning accuracy and fast attitude maneuverability. Because the three experimental areas in Caofeidian Port are all artificial coastlines, and their positions are stable for a long time without being affected by tides, the GF-2 image acquired on May 31, 2015 was used as a reference image. However, there are sandy coasts and bedrock coasts in the coastal area around Xiamen-Quanzhou, which are easily affected by sea tides. Therefore, the Gaofen-2 image close to the acquisition time of the experimental area was selected as a reference image to eliminate or reduce the influence of tides on the accuracy of the results. Since the spatial resolution of the GF-2 image is 1 m/pixel, the uncertainty of the coastline reference position is estimated to be ±1.5 m. Table 1 lists more information about GF-2 images.

1. Image preprocessing. Due to the sensor itself, the atmosphere, terrain and other reasons, the remote sensing satellite will produce errors in the data acquisition process, and the image quality will be affected by the errors, thereby affecting the image accuracy. Therefore, it is necessary to analyze the Landsat 8 OLI experimental image and the GF-2 reference image separately before the experiment. Perform image preprocessing. Firstly, the multispectral data and panchromatic data of GF-2 image were orthorectified respectively by using the RPC parameters of GF-2 image data. Since the georeferencing of multispectral and panchromatic data is better after orthorectification of high-resolution data, image registration is unnecessary, and the NearestNeighbor Diffusion (NNDiffuse) pan sharpening algorithm is directly used for image fusion of the GF-2 reference image. The spatial resolution of the fused GF-2 reference image is 1m. Next, the Nearest Neighbor Diffusion (NNDiffuse) pan sharpening algorithm was used to fuse the Landsat 8 OLI experimental images. After fusion, the spatial resolution of the Landsat 8 OLI experimental images was 15m. A smaller sampling interval can obtain more pixels on the coastline of the same length, and more preliminary coastline pixels are beneficial to the calculation of the gradient and the accuracy of the sub-pixel positioning results, because the smaller the neighborhood window area, the The more in line with the proximity principle of ground objects. Then, the Landsat8 OLI experiment image is radiometrically calibrated by radiometric calibration parameters to obtain radiance value data. Then use the FLAASH model to perform atmospheric correction on the radiance value data (fast atmospheric correction to the Landsat 8 OLI experiment image) to eliminate the influence of atmospheric scattering on the spectrum.

The present invention makes full use of the spectral features of the multi-spectral remote sensing data to calculate the normalized difference water index (NDWI) and the improved normalized difference water index (MNDWI). The specific expressions are shown in formula 1 and formula 2. Both water indices can suppress image noise, and their image histograms are bimodal histograms, which meet the initial assumptions of subsequent algorithms. Considering the impact of suspended sediment on the shore, the short-wave infrared (SWIR) band was selected as another initial image, because the short-wave infrared (SWIR) has a longer wavelength and can penetrate suspended sediment. These choices are consistent with authors of other similar studies.

NDWI＝(p(Green)-p(NIR))/(p(Green)+p(NIR)) (Formula 1)

MNDWI＝(p(Green)-p(MIR))/(p(Green)+p(MIR)) (Formula 2)

2. Image registration. Calculate the geometric displacement between Landsat 8 fused image and GF-2 fused image (reference image). Since the Landsat 8 image and the GF-2 fusion image (reference image) have different image resolutions, the scale-invariant SIFT algorithm is used for image registration to obtain the geometric displacement between the Landsat 8 image and the reference image ( dx, dy).

3. Coastline extraction. The extraction schematic diagram is shown in Fig. 4, because the spectral responses of water and land are different in different bands, the histograms of the combined images of three different bands present bimodality, which meets the initial assumption of the OTSU algorithm, so the present invention adopts the maximum between-class variance method (OTSU ) for image segmentation. The OTSU algorithm divides the original image into two parts, the foreground and the background, by calculating the optimal threshold. Assuming the threshold is T, the optimal threshold T* can be obtained by the following algorithm:

W ₀ =N ₀ /(N ₀ +N ₁ )

W ₁ =N ₁ /(N ₀ +N ₁ )

W ₀ +W ₁ =1

U＝W ₀ *U ₀ +W ₁ *U ₁

G＝W ₀ *(U ₀ -U) ² +W ₁ *(U ₁ -U) ²

Among them, the number of pixels whose gray value is less than the threshold T is N ₀ , the number of pixels whose gray value is greater than the threshold T is N ₁ , W ₀ is the proportion of the number of foreground pixels in the image, and U ₀ is the foreground The average gray level, W ₁ is the proportion of the number of background pixels in the image, U ₁ is the average gray level of the background, U is the average gray level of the whole image, and G is the variance between classes.

When the inter-class variance G is the largest, the threshold T at this time is the optimal threshold T*. Because the greater the inter-class variance between the background and the foreground, the greater the difference between the two parts of the image. When part of the foreground is misclassified as the background or part of the background is misclassified as the foreground, the difference between the two parts will become smaller. Therefore, the optimal threshold T* that maximizes the variance between classes can obtain the image segmentation result with the smallest probability of misclassification.

Then, the segmentation result is processed by region growing algorithm and morphological expansion filter to remove the "impurity" in the land and water, and finally obtain a series of pixels representing the initial coastline position.

4. Coastline change control point extraction. The coastline change control points are extracted from the initial coastline to obtain the initial coastline change control points, and the initial coastline is segmented through the initial coastline change control points to obtain several segmented coastlines. The initial coastline change control point is extracted using the Harris corner detection algorithm, that is, the Harris response value R of each pixel on the initial coastline is calculated in 3; after threshold determination and neighborhood suppression screening, combined with the overall shape of the initial coastline to select The local maximum in R is the change control point on the initial coastline. When calculating the Harris response value R, the pixels in the initial coastline are calculated in sequence according to the connection relationship, so the obtained change control point positions are also stored in sequence according to the connection relationship. The schematic diagram of coastline change control point extraction is shown in Figure 5.

5. Sub-pixel positioning and deviation correction. On the basis of 4, perform sub-pixel positioning on the initial coast points (pixel level) in each segmented coastline. The sub-pixel positioning algorithm based on the neighborhood pixel gray is based on an assumption: according to the principle of proximity, the gray value of each pixel is weighted by the gray value of different objects in the pixel, and the probability of the same object between adjacent pixels maximum. There are only two kinds of features around the preliminary coastline pixels—land and water, so we can think that the gray value of each preliminary coastline pixel is weighted by adjacent pure pixels. Therefore, the sub-pixel localization algorithm is divided into two stages—determining the size of the neighborhood window and calculating the sub-pixel coordinate position. The results of sub-pixel positioning are shown in Figure 6.

The choice of window size needs to meet a prerequisite, that is, on the basis of not violating the principle of proximity, it is guaranteed that the fitting curve can enter from one side of the neighborhood window and pass through from the other side. Therefore, on the basis of the waterway binary image segmented in the initial image, the edge direction of each initial coastline pixel is calculated, and different neighborhood windows are given to the initial coastline pixel according to different edge directions. When the absolute value of the edge direction is less than or equal to 1, a 5×3 neighborhood window is used; when the absolute value of the edge direction is greater than 1, a 3×5 neighborhood window is used.

After confirmation in the neighborhood window, calculate the sub-pixel coordinate position. We describe the boundary of the domain window as a curve through the fitting function, and determine the coordinate position of the sub-pixel anchor point by calculating the parameters of the fitting function. Based on the geometric displacement (dx, dy) between the Landsat 8 fusion image and the reference image calculated in 2, the geographic coordinates of the sub-pixel coastline positioning points are corrected for errors.

6. Anchor point fitting. Since most of the coastlines are basically a smooth line except for the corner points (coastline change control points), the quadratic fitting of the least squares method can be performed on the sub-pixel-level positioning points in each section of the coastline, (ie Use the overall shape of the coastline at the pixel level to correct the positioning results at the sub-pixel level). The sub-pixel anchor points extracted in each segmented coastline in 5 are regarded as a point set, so we can think that the curvature of the sub-pixel anchor points in the same point set does not have a big change, and then the same point set The least squares fitting is carried out on the sub-pixel positioning points in , and the schematic diagram of segment fitting is shown in Figure 7. Connectivity constraints are set at the joints of adjacent line segments and smoothed to obtain a complete and continuous sub-pixel-level coastline to achieve super-resolution mapping. The results of experimental area 1 are shown in Figure 8, and the results of experimental area 2 are shown in Figure 9.

7. Error analysis. Calculate respectively the location error between the coastline sub-pixel positioning coordinates and the reference coastline, the coastline vector result and the reference coastline, and analyze the error results. The error analysis results are shown in Table 2 and Table 3.

Table 2 Error statistics

Table 3 Error statistics

Table 2 summarizes the error statistics (mean error, standard deviation) for each date, each experimental region, and each data type. The average error is obtained by averaging all the errors, and since all errors are obtained by calculating the absolute value of the distance from the final shoreline point to the reference shoreline, the average error is used to account for the level of deviation from the reference shoreline. The standard deviation (STDEV) represents the variability around the mean error.

It can be seen from Table 3 that the sub-pixel positioning algorithm based on the local neighborhood gray can effectively improve the accuracy of coastline extraction. Between 30% and 60%. Comparing sub-pixel coastline accuracies for three different band combinations (selections). Obviously, the coastline obtained with MNDWI data has the highest accuracy, and its total average error value is the smallest. In different fields, most of the average errors are also the smallest. This shows that MNDWI is specifically the water body index with the best noise immunity. Combining the results of Experimental Area 1 and Experimental Area 2, the semi-global super-resolution mapping method proposed by the present invention has good universality and can adapt to coastlines with different curvatures.

Embodiments of the present invention have been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementations, and the above-mentioned specific implementations are only illustrative, rather than restrictive, and those of ordinary skill in the art will Under the enlightenment of the present invention, without departing from the gist of the present invention and the scope of protection of the claims, several improvements and modifications can also be made, and these all belong to the protection of the present invention.

Claims (8)

1. a kind of coastline Super-resolution Mapping based on half global optimization, which is characterized in that include the following steps：

S1,8 land imager images of Landsat and GF-2 are obtained with reference to image, respectively to 8 land imager shadows of Landsat Picture and GF-2 carry out image preprocessing with reference to image, and it is corresponding to obtain 8 fusion evaluations of Landsat, 8 fusion evaluations of Landsat Water body index gray-scale map and GF-2 fusion evaluations；

S2, using GF-2 fusion evaluations as image is referred to, match with reference to image and 8 fusion evaluations of Landsat to described Standard is obtained with reference to the offset parameter between 8 fusion evaluation of image and Landsat；

S3, pixel grade tidal saltmarsh is carried out to the water body index gray-scale map, obtains initial coastline；To being carried out with reference to image Tidal saltmarsh obtains and refers to coastline；

S4, coastline change Control point extraction is carried out to the initial coastline, obtains initial coastline change control point, passes through The initial coastline is segmented by initial coastline change control point, obtains several sections of segmentation coastlines；

S5, the positioning of the sub-pixed mapping based on regional area is carried out to each pixel point in each section of coastline, obtain each section The sub-pixed mapping elements of a fix each put in coastline；The sub-pixed mapping of each point is positioned by the offset parameter described in step S2 and is sat Mark carries out bias correcting；

S6, coastline least square is carried out to the sub-pixed mapping elements of a fix that the bias correcting of all the points in same section of coastline is crossed Fitting, is fitted to a smooth curved section；All curved sections are combined to obtain complete seashore line vector knot Fruit completes the superresolution mapping in coastline.

2. a kind of coastline Super-resolution Mapping based on half global optimization according to claim 1, which is characterized in that Further include：Calculate separately the coastline sub-pixed mapping elements of a fix and with reference between coastline, coastline vector result and reference Site error between coastline and analytical error result.

3. the coastline Super-resolution Mapping based on half global optimization according to claim 1, which is characterized in that step Sub-pixed mapping positioning based on regional area described in S5 comprises the steps of：

S51, the edge direction for calculating each pixel point in each section of initial coastline, according to different edge directions Determine the different neighborhood window of each pixel point of initial coastline；

S52, by fitting function, the field window edge is described as a curve, the parameter of digital simulation function determines The coordinate of the sub-pixed mapping anchor point of regional area.

4. the coastline Super-resolution Mapping based on half global optimization according to claim 1, which is characterized in that step Coastline least square fitting described in S6 comprises the steps of：

S61, it detects and extracts the sub-pixed mapping elements of a fix on each section of coastline；

S62, the sub-pixed mapping elements of a fix are divided into different point sets, the sub-pixed mapping elements of a fix that same point is concentrated is carried out most Small two multiply fitting, obtain sub-pixed mapping grade coastline.

5. a kind of coastline superresolution mapping system based on half global optimization, which is characterized in that including following module：

Image preprocessing image co-registration module refers to image, respectively for obtaining 8 land imager images of Landsat and GF-2 Image preprocessing is carried out with reference to image to 8 land imager images of Landsat and GF-2, obtain 8 fusion evaluations of Landsat, The corresponding water body index gray-scale map of 8 fusion evaluations of Landsat and GF-2 fusion evaluations；

Image registration module, for using GF-2 fusion evaluations as image is referred to, melting with reference to image and Landsat 8 to described Group photo is obtained as being registrated with reference to the offset parameter between 8 fusion evaluation of image and Landsat；

Tidal saltmarsh module obtains initial seashore for carrying out pixel grade tidal saltmarsh to the water body index gray-scale map Line；To carrying out tidal saltmarsh with reference to image, obtains and refer to coastline；

Coastline change Control point extraction obtains just for carrying out coastline change Control point extraction to the initial coastline The initial coastline is segmented by initial coastline change control point, obtains several sections points by beginning coastline change control point Section coastline；

Sub-pixed mapping locating module, for carrying out the sub-pixed mapping based on regional area to each pixel point in each section of coastline Positioning obtains the sub-pixed mapping elements of a fix each put in each section of coastline；Pass through the offset parameter described in Image registration module Bias correcting is carried out to the sub-pixed mapping elements of a fix of each point；

Anchor point fitting module, the sub-pixed mapping elements of a fix crossed for the bias correcting to all the points in same section of coastline into Row coastline least square fitting is fitted to a smooth curved section；All curved sections are combined to have obtained Whole coastline vector result, completes the superresolution mapping in coastline.

6. a kind of coastline superresolution mapping system based on half global optimization according to claim 5, which is characterized in that It further include error analysis module：For calculate separately the coastline sub-pixed mapping elements of a fix and with reference between coastline, seashore Site error between line vector result and reference coastline and analytical error result.

7. the coastline superresolution mapping system based on half global optimization according to claim 5, which is characterized in that sub- picture The sub-pixed mapping positioning based on regional area is comprising with lower module described in first locating module：

Neighborhood window determining module, the edge direction for calculating each pixel point in described each section initial coastline, The different neighborhood window of each pixel point of initial coastline is determined according to different edge directions；

Sub-pixed mapping anchor point coordinate obtaining module, for by fitting function, the field window edge to be described as a song Line, the parameter of digital simulation function determine the coordinate of the sub-pixed mapping anchor point of regional area.

8. the coastline superresolution mapping system based on half global optimization according to claim 5, which is characterized in that positioning Coastline least square fitting described in point fitting module includes with lower module：

Sub-pixed mapping elements of a fix module is extracted, for detecting and extracting the sub-pixed mapping elements of a fix on each section of coastline；

Sub-pixed mapping grade coastline acquisition module, for the sub-pixed mapping elements of a fix to be divided into different point sets, to same point set In the sub-pixed mapping elements of a fix carry out least square fitting, obtain sub-pixed mapping grade coastline.