A mountain river extraction method and system

Abstract

The present invention relates to a mountain river extraction method and system, the method comprising: extracting hydrological features from DEM data to obtain a micro-watershed; pre-processing a multi-spectral image to obtain a water index map and a linear enhancement map; in the micro-watershed, determining potential river pixels according to the linear enhancement map, and assigning values ​​to the potential river pixels; determining a non-water body threshold for using the image based on the water index map; using a regional growing method to determine a river pixel set based on a non-water body threshold and a water body threshold; disconnecting and binarizing the river pixels in the river pixel set to obtain a mountain river extraction map. The present invention effectively combines the hydrological features extracted based on DEM data with the water features obtained from the multi-spectral remote sensing image, solves the problem of difficulty in formulating a unified rule caused by the huge differences in spectral features of mountain rivers encountered in river extraction using remote sensing images, and improves the accuracy of mountain river extraction.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method and system for extracting rivers in mountainous areas. Background Art

[0002] As an element of the earth's hydrosphere, rivers are the main mode of terrestrial water circulation. They play an important role in the transportation and redistribution of global materials and energy, and play an incomparable ecological, economic and social service function. With the rapid development of remote sensing technology in recent years, river extraction based on remote sensing images has replaced traditional field measurements and become the main way to obtain river-related information. Compared with plain rivers, mountain rivers are more difficult to automatically extract because they have more terminal tributaries, more dramatic changes in river width, and huge differences in river spectra in different regions.

[0003] In order to improve the accuracy of automatic extraction of mountain rivers, many scholars have tried it. For example, studies such as "An Automated Method for Extracting Rivers and Lakes from Landsat Imagery" and "River Delineation from Remotely Sensed Imagery Using a Multi-Scale Classification Approach" use linear enhancement to enhance the contrast between rivers and backgrounds in spectral images, reduce the differences between rivers in different regions, widths, and water quality, and make them easier to extract by unified standards. "River Detection in Remotely Sensed Imagery Using Gabor Filtering and Path Opening" uses the connectivity of rivers to obtain a complete river network using path opening operations. Among the commonly used mountain river features, linear features and connectivity features both rely on the spectral features of remote sensing images, but the spectral features of mountain rivers are not stable enough, so the problem of difficulty in setting unified rules is only alleviated but not solved. It is well known that compared with plain rivers, mountain rivers are more dependent on terrain, and the accuracy of DEM river extraction results based on D8 or multi-flow direction algorithms is higher than that in plain areas, and there are no problems such as terminal tributaries cannot be extracted, line breaks, and topological errors. However, the existing publicly available DEM data still have problems such as low precision and lack of timeliness. In addition, due to the limitations of algorithms and data organization, the extraction results cannot clearly indicate whether there is actually a water body, so there is a problem of low extraction accuracy. Summary of the invention

[0004] The purpose of the present invention is to provide a mountain river extraction method and system to improve the accuracy of mountain river extraction.

[0005] To achieve the above object, the present invention provides a method for extracting mountain rivers, the method comprising:

[0006] Step S1: Obtain multispectral images and DEM data corresponding to the area to be extracted;

[0007] Step S2: extracting hydrological features from the DEM data to obtain a micro-watershed;

[0008] Step S3: preprocessing the multispectral image to obtain a water index map and a linear enhancement map;

[0009] Step S4: in the micro-watershed, determining potential river pixels according to the linear enhancement map, and assigning values ​​to the potential river pixels;

[0010] Step S5: determining a non-water body threshold using an image based on the water body index map;

[0011] Step S6: using a region growing method to determine a river pixel set based on the non-water body threshold;

[0012] Step S7: disconnecting and binarizing the river pixels in the river pixel set to obtain a mountain river extraction map.

[0013] Optionally, extracting hydrological features from the DEM data to obtain a micro-watershed specifically includes:

[0014] Step S21: performing depression filling processing on the DEM data to obtain the pre-processed DEM data;

[0015] Step S22: extracting river hydrological features from the pre-processed DEM data to obtain a flow direction grid;

[0016] Step S23: determining a flow grid according to the flow direction grid;

[0017] Step S24: extracting the preprocessed DEM data according to the first threshold value to obtain a river linear grid;

[0018] Step S25: constructing a buffer zone corresponding to the river linear grid;

[0019] Step S26: Processing the flow grid according to the second threshold value to obtain a plurality of small watersheds;

[0020] Step S27: Overlay analysis is performed on each of the small watersheds and the river linear grid of the constructed buffer zone to generate a micro-watershed.

[0021] Optionally, the preprocessing of the multispectral image to obtain a water index map and a line enhancement map specifically includes:

[0022] Step S31: performing cropping and cloud removal processing on the multispectral image to obtain the multispectral image after cloud removal;

[0023] Step S32: extracting the multispectral image after cloud removal using a normalized water index to obtain a water index map;

[0024] Step S33: performing linear enhancement on the rivers in the water index map to obtain a linear enhancement map.

[0025] Optionally, the determining of the non-water body threshold of the image based on the water body index map specifically includes:

[0026] Step S51: constructing an image histogram based on the water index map, and finding the optimal threshold using the OTSU method;

[0027] Step S52: using an iterative method, starting from the optimal threshold value, searching for water body index values ​​with the largest change rate on both sides of the histogram as the non-water body threshold value and the water body threshold value of the image.

[0028] Optionally, the adopting of a region growing method to determine a river pixel set based on the non-water body threshold specifically includes:

[0029] Step S61: taking river pixels in the micro-watershed whose linear enhancement value is greater than 0.4 and whose water index value is greater than the non-water body threshold as initial seed points;

[0030] Step S62: taking the initial seed point as the starting point;

[0031] Step S63: determine whether similar pixels corresponding to the initial seed point are found; if similar pixels corresponding to the initial seed point are found, the similar pixels are included in the river pixel set; if similar pixels corresponding to the initial seed point are not found, the process ends; the seed point neighborhood pixels in the micro-watershed, whose linear enhancement value is greater than 0.2 and whose water body index value is greater than the non-water body threshold, are regarded as similar pixels;

[0032] Step S64: merge the initial seed point and the similar pixel to form a new initial seed point, and return to "Step S63".

[0033] The present invention also provides a mountain river extraction system, the system comprising:

[0034] The acquisition module is used to obtain the multispectral image and DEM data corresponding to the area to be extracted;

[0035] A micro-watershed determination module is used to extract hydrological features from the DEM data to obtain a micro-watershed;

[0036] A preprocessing module, used for preprocessing the multispectral image to obtain a water index map and a linear enhancement map;

[0037] A river pixel assignment module, used for determining potential river pixels in the micro-watershed according to the linear enhancement map, and assigning values ​​to the potential river pixels;

[0038] A threshold determination module, used to determine a non-water body threshold using an image based on the water body index map;

[0039] A river pixel set determination module, used for determining a river pixel set based on the non-water body threshold by adopting a region growing method;

[0040] The mountain river extraction module is used to connect the river pixels in the river pixel set by broken lines and perform binarization processing to obtain a mountain river extraction map.

[0041] Optionally, the micro-flow area determination module specifically includes:

[0042] A first preprocessing unit is used to perform depression filling processing on the DEM data to obtain the preprocessed DEM data;

[0043] A flow direction grid determination unit is used to extract river hydrological characteristics from the pre-processed DEM data to obtain a flow direction grid;

[0044] A sink amount grid determination unit, used for determining a sink amount grid according to the flow direction grid;

[0045] A river linear grid determination unit, used for extracting the pre-processed DEM data according to a first threshold value to obtain a river linear grid;

[0046] A buffer zone construction unit, used to construct a buffer zone corresponding to the river linear grid;

[0047] A small watershed determination unit is used to process the confluence grid according to a second threshold value to obtain a plurality of small watersheds;

[0048] The micro-watershed determination unit is used to perform overlay analysis on each of the small watersheds and the river linear grid of the constructed buffer zone to generate a micro-watershed.

[0049] Optionally, the preprocessing module specifically includes:

[0050] A second pre-processing unit is used to perform cropping and cloud removal processing on the multispectral image to obtain the multispectral image after cloud removal;

[0051] A water index map determining unit, configured to extract the multispectral image after cloud removal using a normalized water index to obtain a water index map;

[0052] The linear enhancement map determining unit is used to perform linear enhancement on the rivers in the water body index map to obtain a linear enhancement map.

[0053] Optionally, the threshold determination module specifically includes:

[0054] An optimal threshold determination unit, used to construct an image histogram based on the water index map and find the optimal threshold using the OTSU method;

[0055] The non-water body threshold and water body threshold determination unit is used to iteratively search for the water body index value with the largest change rate from the optimal threshold to both sides of the histogram as the non-water body threshold and water body threshold of the image.

[0056] Optionally, the river pixel set determination module specifically includes:

[0057] The initial seed point determination unit is used to take river pixels in the micro-watershed whose linear enhancement value is greater than 0.4 and whose water body index value is greater than the non-water body threshold as initial seed points;

[0058] A starting point determination unit, used to take the initial seed point as the starting point;

[0059] A judgment unit is used to judge whether similar pixels corresponding to the initial seed point are found; if similar pixels corresponding to the initial seed point are found, the similar pixels are included in the river pixel set; if similar pixels corresponding to the initial seed point are not found, the process ends; the seed point neighborhood pixels in the micro-watershed, whose linear enhancement value is greater than 0.2 and whose water body index value is greater than the non-water body threshold, are regarded as similar pixels;

[0060] The merging unit is used to merge the initial seed point with the similar pixel to form a new initial seed point, and return it to the "judgment unit".

[0061] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0062] The present invention effectively combines the hydrological features extracted based on DEM data with the water body features obtained from multispectral remote sensing images, thereby solving the problem of difficulty in formulating unified rules due to the huge differences in spectral features of mountain rivers encountered in river extraction using remote sensing images, and improving the accuracy of mountain river extraction. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0064] Figure 1 This is a flow chart of the mountain river extraction method of the present invention;

[0065] Figure 2 This is a structural diagram of the mountain river extraction system of the present invention;

[0066] Figure 3 This is a flow chart of mountain river extraction combining DEM and multispectral remote sensing images in the present invention;

[0067] Figure 4 It is a schematic diagram of the micro-watershed of the present invention;

[0068] Figure 5 This is a schematic diagram of the MNDWI index results of the present invention.

[0069] Figure 6 Schematic diagram of the LFE operator of the present invention;

[0070] Figure 7 This is the LCLFE linear enhancement diagram of the present invention;

[0071] Figure 8 It is the LCLFE line enhancement diagram after hydrological reconstruction of the present invention;

[0072] Fig. 9 The present invention is a schematic diagram of determining a non-water body threshold and a water body threshold based on a grayscale histogram;

[0073] Fig.10 This is a schematic diagram of the mountain river extraction results and real data of the present invention. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] The purpose of the present invention is to provide a mountain river extraction method and system to improve the accuracy of mountain river extraction.

[0076] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0077] Example 1

[0078] like Figure 1 As shown, the present invention discloses a method for extracting mountain rivers, the method comprising:

[0079] Step S1: Obtain multispectral images and DEM data corresponding to the area to be extracted.

[0080] Step S2: extracting hydrological features from the DEM data to obtain micro-watersheds.

[0081] Step S3: pre-processing the multispectral image to obtain a water index map and a line enhancement map.

[0082] Step S4: In the micro-watershed, potential river pixels are determined according to the linear enhancement map, and values ​​are assigned to the potential river pixels.

[0083] Step S5: Determine the non-water body threshold using the image based on the water body index map.

[0084] Step S6: using the region growing method to determine the river pixel set based on the non-water body threshold.

[0085] Step S7: disconnecting and binarizing the river pixels in the river pixel set to obtain a mountain river extraction map.

[0086] The following is a detailed discussion of each step:

[0087] In this embodiment, Landsat 8OLI is used as an example of multispectral imagery, and SRTM is used as an example of DEM data.

[0088] Step S2: extracting hydrological features from the DEM data to obtain micro-watersheds, specifically including:

[0089] Step S21: performing depression-filling processing on the DEM data to obtain the pre-processed DEM data.

[0090] Step S22: extracting river hydrological features from the pre-processed DEM data to obtain a flow direction grid.

[0091] Step S23: determining a flow grid according to the flow direction grid.

[0092] Step S24: extracting the preprocessed DEM data according to the first threshold value to obtain a river linear grid; the river grid presented in the form of a line is referred to as the river linear grid; one percent of the basin area is referred to as the first threshold value.

[0093] Step S25: constructing a buffer zone corresponding to the river linear grid.

[0094] Step S26: Processing the flow grid according to the second threshold value to obtain a plurality of small watersheds.

[0095] Step S27: Overlay analysis is performed on each of the small watersheds and the river linear grid of the constructed buffer zone to generate a micro-watershed.

[0096] Step S3: preprocessing the multispectral image to obtain a water index map and a linear enhancement map, specifically including:

[0097] Step S31: performing cropping and cloud removal processing on the multispectral image to obtain the multispectral image after cloud removal.

[0098] Step S32: extract the multispectral image after de-clouding using the normalized water index to obtain a water index map; in this embodiment, extract the multispectral image after de-clouding using MNDWI, NDWI and AWEI.

[0099] Step S33: Linearly enhance the rivers in the water index map to obtain a linearly enhanced map. Specifically, in this embodiment, linearly enhance the rivers in the water index map using LCLFE filtering, Gabor filtering, and Frangi filtering.

[0100] Step S4: In the micro-watershed, potential river pixels are determined according to the linear enhancement map, and values ​​are assigned to the potential river pixels; specifically, the river pixels in the micro-watershed and in the top three linear enhancement values ​​in the 8-neighborhood of the linear enhancement map are taken as potential river pixels.

[0101] The potential river pixels are assigned values, and the specific formula is:

[0102]

[0103] Among them, HI i represents the value assigned to the ith potential river pixel, α and ε are both set constants, and acc i It represents the grid value of the flow rate corresponding to the linear river pixel closest to the potential river pixel i (the grid value of the linear river is 1), accmax represents the maximum flow rate in the basin, and t is the level of the last tributary in the basin. i The value range is The LCLFE value of potential river pixel i that is out of the value range will not be corrected.

[0104] If the present invention does not perform value assignment processing on potential river pixels, they may be missed in the subsequent region growing process, thereby affecting the subsequent mountain river extraction accuracy.

[0105] Step S5: Determine the non-water body threshold of the image based on the water body index map, specifically including:

[0106] Step S51: construct an image histogram based on the water index map, and use the OTSU method to find the optimal threshold.

[0107] Step S52: using an iterative method, starting from the optimal threshold value, searching for water body index values ​​with the largest change rate on both sides of the histogram as non-water body threshold values ​​for use in the image.

[0108] Step S6: using a region growing method to determine a river pixel set based on the non-water body threshold, specifically including:

[0109] Step S61: The river pixels in the micro-watershed whose linear enhancement value is greater than 0.4 and whose water index value is greater than the non-water body threshold are taken as initial seed points.

[0110] Step S62: taking the initial seed point as the starting point.

[0111] Step S63: Determine whether similar pixels corresponding to the initial seed point are found; if similar pixels corresponding to the initial seed point are found, the similar pixels are included in the river pixel set; if similar pixels corresponding to the initial seed point are not found, the process ends. In this embodiment, river pixels in the neighborhood of the seed point that are in the micro-watershed and have a linear enhancement value (i.e., LCLFE value) greater than 0.2 and a water body index value (i.e., MNDWI value) greater than the non-water body threshold are regarded as similar pixels.

[0112] Step S64: merge the initial seed point and the similar pixel to form a new initial seed point, and return to "Step S63".

[0113] Step S65: Eliminate the links in the river pixel set whose number of connected pixels is less than the set number of pixels.

[0114] Example 2

[0115] like Figure 2 As shown, the present invention also discloses a mountain river extraction system, the system comprising:

[0116] The acquisition module 201 is used to acquire the multispectral image and DEM data corresponding to the area to be extracted.

[0117] The micro-watershed determination module 202 is used to extract hydrological features from the DEM data to obtain micro-watersheds.

[0118] The preprocessing module 203 is used to preprocess the multispectral image to obtain a water index map and a line enhancement map.

[0119] The river pixel assignment module 204 is used to determine potential river pixels in the micro-watershed according to the linear enhancement map and assign values ​​to the potential river pixels.

[0120] The threshold determination module 205 is used to determine the non-water body threshold using the image based on the water body index map.

[0121] The river pixel set determination module 206 is used to determine the river pixel set based on the non-water body threshold by using a region growing method.

[0122] The mountain river extraction module 207 is used to connect the river pixels in the river pixel set by broken lines and perform binarization processing to obtain a mountain river extraction map.

[0123] The following is a detailed discussion of each module:

[0124] As an optional implementation manner, the micro-flow area determination module 202 of the present invention specifically includes:

[0125] The first preprocessing unit is used to perform depression filling processing on the DEM data to obtain the preprocessed DEM data.

[0126] The flow direction grid determination unit is used to extract river hydrological characteristics from the pre-processed DEM data to obtain a flow direction grid.

[0127] The confluence grid determining unit is used to determine the confluence grid according to the flow direction grid.

[0128] The river linear grid determination unit is used to extract the pre-processed DEM data according to a first threshold value to obtain a river linear grid.

[0129] The buffer zone construction unit is used to construct a buffer zone corresponding to the river linear grid.

[0130] The small watershed determination unit is used to process the confluence grid according to a second threshold value to obtain a plurality of small watersheds.

[0131] The micro-watershed determination unit is used to perform overlay analysis on each of the small watersheds and the river linear grid of the constructed buffer zone to generate a micro-watershed.

[0132] As an optional implementation, the preprocessing module 203 of the present invention specifically includes:

[0133] The second preprocessing unit is used to perform cropping and cloud removal on the multispectral image to obtain the multispectral image after cloud removal.

[0134] The water index map determining unit is used to extract the multispectral image after cloud removal by using the normalized water index to obtain a water index map.

[0135] The linear enhancement map determining unit is used to perform linear enhancement on the rivers in the water body index map to obtain a linear enhancement map.

[0136] As an optional implementation, the threshold determination module 205 of the present invention specifically includes:

[0137] The optimal threshold determination unit is used to construct an image histogram based on the water body index map and find the optimal threshold using the OTSU method.

[0138] The non-water body threshold and water body threshold determination unit is used to iteratively search for the water body index value with the largest change rate from the optimal threshold to both sides of the histogram as the non-water body threshold and water body threshold of the image.

[0139] As an optional implementation manner, the river pixel set determination module 206 of the present invention specifically includes:

[0140] The initial seed point determination unit is used to take river pixels in the micro-watershed whose linear enhancement value is greater than 0.4 and whose water body index value is greater than the non-water body threshold as initial seed points.

[0141] A starting point determination unit is used to take the initial seed point as the starting point.

[0142] The judgment unit is used to judge whether similar pixels corresponding to the initial seed point are found; if similar pixels corresponding to the initial seed point are found, the similar pixels are included in the river pixel set; if similar pixels corresponding to the initial seed point are not found, the process ends; the seed point neighborhood pixels that are in the microwatershed and have a linear enhancement value greater than 0.2 and a water body index value greater than the non-water body threshold are regarded as similar pixels.

[0143] The merging unit is used to merge the initial seed point with the similar pixel to form a new initial seed point, and return it to the "judgment unit".

[0144] The river pixel set determination module 206 further includes:

[0145] The elimination unit is used to eliminate the links whose number of connected pixels in the river pixel set is less than the set number of pixels.

[0146] The present invention combines the hydrological features extracted from DEM data with the water body features obtained from multispectral remote sensing images to realize the automatic extraction of mountain rivers, solves the problem of difficulty in formulating unified rules caused by the huge differences in spectral features of rivers in different regions encountered when using remote sensing images for mountain river extraction, and provides corresponding ideas for the automatic extraction of mountain rivers.

[0147] The present invention effectively combines the hydrological features extracted based on DEM data with the water body features obtained from multispectral remote sensing images, solves the problem of difficulty in formulating unified rules caused by the huge differences in spectral features of mountain rivers encountered in river extraction using remote sensing images, and provides corresponding ideas for the automatic extraction of mountain rivers. The main parts are all implemented by Python programming, reducing manual participation. Compared with the prior art, the present invention has the following beneficial technical effects: First, the image is subjected to length-based linear enhancement filtering, which reduces the interference of short-line objects other than rivers and can enhance long connected objects. Second, the river hydrological features extracted from DEM are used instead of direct DEM data, which avoids the errors caused by the data source differences between DEM data and multispectral remote sensing images. Third, in view of the problem that it is difficult to set unified rules due to the spectral differences of mountain rivers in multispectral images, the hydrological features obtained by DEM data are used to reassign pixels that meet the river characteristics to meet the unified requirements. Fourth, the present invention has a good effect on mountain river extraction and higher robustness.

[0148] Example 3

[0149] This example uses the extraction method in Example 1 for discussion. Figure 3 , specifically including the following steps:

[0150] The first step is to download the Landsat 8OLI image (30m spatial resolution) and SRTM (30m spatial resolution) DEM data covering the study area and perform corresponding preprocessing respectively: the Landsat 8OLI image is de-clouded by masking the pixels marked as clouds and cloud shadows in its qa band, and the hydrological module of ArcGIS is used to remove the depressions in the DEM data.

[0151] Step 2: Through the hydrology module of ArcGIS, the D8 algorithm (explanation of terms is shown in Table 1) is used to calculate the flow direction grid and confluence grid of the river, and the river linear grid is obtained with 1% of the basin area as the default threshold (i.e., the first threshold).

[0152] Step 3: Set 0.09km based on the flow grid generated in step 2 2 The threshold (second threshold) is used to generate a large number of small watersheds. A 500m buffer zone is created for the river linear grid, and the small watersheds and their boundaries that intersect with the buffer zone are retained. The final generated range is the river micro-watershed. The specific range of the micro-watershed can be seen in Figure 4 , which narrows the research scope and can remove the interference of a large number of irrelevant areas on the extraction results.

[0153] Step 4: Use Python to read remote sensing images, normalize the green band and shortwave infrared band, and obtain the MNDWI index map using formula (2) to highlight the difference between water bodies and background.

[0154]

[0155] Among them, GREEN is the green band and SWIR is the short-wave infrared band. Figure 5 , which effectively realizes the dimensionality reduction of multispectral images and the contrast stretching of water and non-water bodies.

[0156] Step 5: If Figure 6 As shown in Figure 1, four LFE operators and formula (3) are used to implement LFE filtering. The maximum result value of operators in different directions is taken as the LFE response result of the pixel. Then, the simple connectivity of the results is achieved through double threshold tracking. The strong and weak thresholds are set to 0.4 and 0.2 respectively.

[0157]

[0158] Among them, the LFE operator and the dual threshold tracking method are existing algorithms.

[0159] Then, the result is divided into blocks in 8-neighborhoods using a thinning algorithm, and the LCLFE index value is constructed according to formula (4):

[0160]

[0161] Among them, LFE max Represents the highest LFE index in the same pixel block, LFE min Represents the minimum LFE index in the block, and num represents the number of pixels in the block. Figure 7 The LCLFE results are shown, and it can be clearly seen that it suppresses short connected features and strengthens long connected linear features.

[0162] Step 6: Identify the top three pixels in the 8-neighborhood for all pixels located in the microwatershed and regard them as potential river pixels.

[0163] Step 7: Re-assign the LCLFE value of potential river pixel i according to formula (5):

[0164]

[0165] Among them, HI i represents the value assigned to the ith potential river pixel, α and ε are both set constants, and acci It represents the grid value of the confluence corresponding to the linear river pixel closest to the potential river pixel i (the grid value of the linear river is 1), accmax represents the maximum confluence value in the basin, and t is the level of the last tributary in the basin. The specific results after the assignment can be seen in Figure 8 , some weakly responding river pixels have also been strengthened accordingly.

[0166] Step 8: Construct an image histogram for the MNDWI image. First, use the OTSU method to find the optimal threshold. Then, use an iterative method to search from the optimal threshold to both sides of the histogram to obtain the water index value with the maximum change rate, and determine the non-water body and water body thresholds of the image. For details, see Fig. 9 .

[0167] Step 9: Construct regional growth rules, taking pixels with LCLFE values ​​greater than 0.4 and MNDWI values ​​greater than the non-water body threshold in the micro-watershed as initial seed points, and constructing them into a river pixel set. When the LCLFE value of the neighboring pixel is greater than 0.2 and it is in the micro-watershed, and the pixel with MNDWI value greater than the non-water body threshold is included in the river pixel set.

[0168] Step 10: Repeat the growth rule in step 9 until no new pixels are included in the set and remove pixel blocks with less than 20 connected pixels.

[0169] Step 11: Inverse normalize the LCLFE result value to generate a cost grid, and set the cost of the extracted river pixels to 0, the cost of pixels less than the non-water body threshold and non-micro-watershed pixels to nan, and the unreachable area. Connect each non-adjacent block to finally obtain the mountain river extraction result, see Fig.10 As shown in (a), the corresponding real result is as follows ( Fig.10 As shown in (b) in Figure 2, it can be found that the two figures are basically consistent.

[0170] Table 1 English abbreviations and definitions of key terms

[0171]

[0172]

[0173] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0174] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for extracting mountain rivers, It is characterized in that The method comprises: Step S1: Obtain multispectral images and DEM data corresponding to the area to be extracted; Step S2: extracting hydrological features from the DEM data to obtain a micro-watershed, including: Step S21: performing depression filling processing on the DEM data to obtain the pre-processed DEM data; Step S22: extracting river hydrological features from the pre-processed DEM data to obtain a flow direction grid; Step S23: determining a flow grid according to the flow direction grid; Step S24: extracting the pre-processed DEM data according to the first threshold value to obtain a river linear grid; Step S25: constructing a buffer zone corresponding to the river linear grid; Step S26: Processing the flow grid according to the second threshold value to obtain a plurality of small watersheds; Step S27: performing an overlay analysis on each of the small watersheds and the river linear grid of the constructed buffer zone to generate a micro-watershed; Step S3: preprocessing the multispectral image to obtain a water index map and a linear enhancement map; Step S4: in the micro-watershed, determining potential river pixels according to the linear enhancement map, and assigning values ​​to the potential river pixels; Step S5: determining a non-water body threshold using an image based on the water body index map; Step S6: using a region growing method to determine a river pixel set based on the non-water body threshold, including: Step S61: taking river pixels in the micro-watershed whose linear enhancement value is greater than 0.4 and whose water index value is greater than the non-water body threshold as initial seed points; Step S62: taking the initial seed point as the starting point; Step S63: determine whether similar pixels corresponding to the initial seed point are found; if similar pixels corresponding to the initial seed point are found, the similar pixels are included in the river pixel set; if similar pixels corresponding to the initial seed point are not found, the process ends; the seed point neighborhood pixels in the micro-watershed, whose linear enhancement value is greater than 0.2 and whose water body index value is greater than the non-water body threshold, are regarded as similar pixels; Step S64: merging the initial seed point with the similar pixel to form a new initial seed point, and returning to step S63; Step S7: disconnecting and binarizing the river pixels in the river pixel set to obtain a mountain river extraction map.

2. The method for extracting mountain rivers according to claim 1, It is characterized in that The preprocessing of the multispectral image to obtain a water index map and a linear enhancement map specifically includes: Step S31: performing cropping and cloud removal processing on the multispectral image to obtain the multispectral image after cloud removal; Step S32: extracting the multispectral image after cloud removal using a normalized water index to obtain a water index map; Step S33: performing linear enhancement on the rivers in the water index map to obtain a linear enhancement map.

3. The mountain river extraction method according to claim 1, It is characterized in that The determining of the non-water body threshold of the image based on the water body index map specifically includes: Step S51: constructing an image histogram based on the water index map, and finding the optimal threshold using the OTSU method; Step S52: using an iterative method, starting from the optimal threshold value, searching for water body index values ​​with the largest change rate on both sides of the histogram as the non-water body threshold value and the water body threshold value of the image.

4. A mountain river extraction system, It is characterized in that The system comprises: The acquisition module is used to obtain the multispectral image and DEM data corresponding to the area to be extracted; A micro-watershed determination module is used to extract hydrological features from the DEM data to obtain a micro-watershed; A preprocessing module, used for preprocessing the multispectral image to obtain a water index map and a linear enhancement map; A river pixel assignment module, used for determining potential river pixels in the micro-watershed according to the linear enhancement map, and assigning values ​​to the potential river pixels; A threshold determination module, used to determine a non-water body threshold using an image based on the water body index map; A river pixel set determination module, used for determining a river pixel set based on the non-water body threshold by adopting a region growing method; A mountain river extraction module, used for connecting the river pixels in the river pixel set by broken lines and performing binarization processing to obtain a mountain river extraction map; The micro-flow area determination module comprises: A first preprocessing unit is used to perform depression filling processing on the DEM data to obtain the preprocessed DEM data; A flow direction grid determination unit is used to extract river hydrological characteristics from the pre-processed DEM data to obtain a flow direction grid; A sink amount grid determination unit, used for determining a sink amount grid according to the flow direction grid; A river linear grid determination unit, used for extracting the pre-processed DEM data according to a first threshold value to obtain a river linear grid; A buffer zone construction unit, used to construct a buffer zone corresponding to the river linear grid; A small watershed determination unit is used to process the confluence grid according to a second threshold value to obtain a plurality of small watersheds; A micro-watershed determination unit is used to perform an overlay analysis on each of the small watersheds and the river linear grid of the constructed buffer zone to generate a micro-watershed; The river pixel set determination module comprises: The initial seed point determination unit is used to take river pixels in the micro-watershed whose linear enhancement value is greater than 0.4 and whose water body index value is greater than the non-water body threshold as initial seed points; A starting point determination unit, used to take the initial seed point as the starting point; A judgment unit is used to judge whether similar pixels corresponding to the initial seed point are found; if similar pixels corresponding to the initial seed point are found, the similar pixels are included in the river pixel set; if similar pixels corresponding to the initial seed point are not found, the process ends; the seed point neighborhood pixels in the micro-watershed, whose linear enhancement value is greater than 0.2 and whose water body index value is greater than the non-water body threshold, are regarded as similar pixels; The merging unit is used to merge the initial seed point with the similar pixel to form a new initial seed point, and return to execute the steps in the judgment unit.

5. The mountain river extraction system according to claim 4, It is characterized in that The preprocessing module specifically includes: A second pre-processing unit is used to perform cropping and cloud removal processing on the multispectral image to obtain the multispectral image after cloud removal; A water index map determining unit, configured to extract the multispectral image after cloud removal using a normalized water index to obtain a water index map; The linear enhancement map determining unit is used to perform linear enhancement on the rivers in the water body index map to obtain a linear enhancement map.

6. The mountain river extraction system according to claim 4, It is characterized in that The threshold determination module specifically includes: An optimal threshold determination unit, used to construct an image histogram based on the water index map and find the optimal threshold using the OTSU method; The non-water body threshold and water body threshold determination unit is used to iteratively search for the water body index value with the largest change rate from the optimal threshold to both sides of the histogram as the non-water body threshold and water body threshold of the image.

Images