A method and system for extracting intertidal flats based on tidal flat index

Abstract

The present invention discloses a method and system for extracting intertidal mudflats based on tidal index, the method comprising: obtaining remote sensing images of a target area within a target period; generating low tide synthetic images and high tide synthetic images based on the remote sensing images by using the quantile synthesis method; calculating the tidal index corresponding to each pixel position by using the low tide and high tide synthetic images; performing object-oriented multi-scale segmentation using the calculation result of the tidal index to obtain a patch object, and then classifying the patch object using a classification model to obtain the distribution result of the intertidal mudflats in the target area. The technical solution of the present invention first generates low tide synthetic images and high tide synthetic images, and then simultaneously incorporates the spectral reflectance of low tide and high tide into the calculation of the tidal index, thereby overcoming the technical obstacles brought by the periodic flooding of tides to the remote sensing extraction of intertidal mudflats, thereby effectively improving the accuracy of mudflat extraction.

Description

Technical Field

[0001] The invention belongs to the technical field of classification extraction of remote sensing images, and in particular relates to a method and system for extracting intertidal mudflats based on tidal flat index. Background Art

[0002] Tidal flats are located in the intertidal zone where the marine environment transitions to the terrestrial environment. They are an important part of the marine, terrestrial and freshwater ecosystems. They play an important ecological and economic role in maintaining coastal stability, preventing wind and sand, maintaining biodiversity and developing human activities. However, due to the impact of climate change and human activities, such as sea level rise, land reclamation for urbanization, coastal overdevelopment, and aquaculture expansion, the area of ​​intertidal flats is decreasing, resulting in serious damage to coastal wetland ecosystems and a sharp decline in biodiversity. Therefore, in the severe situation where coastal tidal flats are increasingly being destroyed, the rapid and accurate extraction of the spatial distribution of intertidal flats and the systematic grasp of the area and distribution of tidal flats can provide a data basis for the scientific management, protection and ecological restoration of coastal wetlands, and are therefore crucial to the sustainable development of the coastal zone.

[0003] In recent years, remote sensing technology has been widely used in the extraction and monitoring of intertidal mudflats due to its advantages of accuracy, real-time, multi-scale, and short recurrence period. At present, the traditional common methods for extracting intertidal mudflats can be summarized into three categories: methods based on tidal models or terrain, machine learning classification methods, and knowledge-based decision tree classification methods. Among them, due to the complex spatial dynamic changes of coastal terrain and tidal water levels, methods based on tidal models or terrain have technical defects such as uncertainty, difficulty in obtaining multi-source data in the early stage, and cumbersome implementation processes; machine learning classification methods need to rely on a large number of sample training, but the selection is labor-intensive, costly, and lacks repeatability; knowledge-based decision tree classification methods need to determine threshold conditions based on experience, but the unified threshold is not suitable for the extraction of large areas of mudflats, and there are technical defects such as more manual intervention and lack of automation.

[0004] Secondly, since the periodic flooding of tides brings great difficulties to the extraction of tidal flats, the existing methods for extracting intertidal tidal flats based on remote sensing images mostly use single-phase images and rely on auxiliary data such as tidal water level, terrain information, and training samples for data sources; in terms of identification methods, most of them rely on thresholds and post-processing of manual intervention. However, there is currently no index specifically used to extract tidal flats based on the characteristics of tidal periodic flooding, and a set of tidal flat extraction methods has been formed with the constructed tidal flat index as the data source. Therefore, the existing methods have the problem that the extraction results of intertidal tidal flats are not accurate enough, and the extraction methods are not universal, intelligent, and reproducible. Summary of the invention

[0005] The purpose of the present invention is to overcome the problem of inaccurate intertidal flat extraction results in the prior art, improve the accuracy of intertidal flat extraction, and especially fill the gap in the prior art of tidal flat extraction technology that takes into account tidal periodic inundation. Furthermore, the technical solution of the present invention provides a method and system for extracting intertidal flats based on tidal flat index, which generates low tide synthetic images and high tide synthetic images, and at the same time incorporates the spectral reflectance of low tide and high tide into the calculation of tidal flat index, overcomes the difficulties brought by tidal periodic inundation to the remote sensing extraction of intertidal flats, solves the problem of over-reliance on manual intervention and difficulty in obtaining auxiliary data, thereby effectively improving the accuracy of tidal flat extraction.

[0006] On the one hand, the present invention provides a method for extracting intertidal mudflats based on tidal flat index, which comprises the following steps:

[0007] Step 1: Acquire remote sensing images of the target area within the target period;

[0008] Step 2: generating a low tide synthetic image and a high tide synthetic image by using a quantile synthesis method based on the remote sensing image;

[0009] Step 3: Calculate the tidal flat index corresponding to each pixel position using the low tide and high tide synthetic images;

[0010] Step 4: Use the calculation result of the tidal flat index to perform object-oriented multi-scale segmentation to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal tidal flats in the target area. Among them, the distribution results of the intertidal tidal flats are actually the intertidal tidal flats in the target area extracted by the technical solution of the present invention.

[0011] The classification model involved in step 4 is set as a remote sensing classification method such as an expert decision tree, a neural network, a support vector machine, and a nearest neighbor algorithm, and a classification model is trained using remote sensing classification methods using training samples of different land types. The input of the classification model is the characteristics of the plaque, and the output result is the type of the plaque; or the input of the classification model is the characteristics of the plaque, and the output result is the classification result of whether the plaque is an intertidal mudflat. In addition, the training samples of the classification model are preferably samples generated by the same type of remote sensing images.

[0012] Further optionally, the calculation formula of the tidal flat index is as follows:

[0013]

[0014] Where TFRI is the tidal flat index, Bandi low is the spectral reflectance of the ith band at a certain pixel position in the low tide synthetic image, Bandj high and Bandk highare the spectral reflectances of the j-th band and the k-th band at the same pixel position in the high tide synthetic image, respectively;

[0015] Among them, the i-th band is a distinctive feature band obtained based on the low-tide synthetic image for identifying the mudflats and other landforms; the j-th and k-th bands are based on the high-tide synthetic image for identifying the mudflats and other landforms.

[0016] Further optionally, the i-th band is the 8th band, the j-th band is the 6th band, and the k-th band is the 7th band.

[0017] Further optionally, the i-th band, the j-th band, and the k-th band are all distinguishing characteristic bands extracted by DNA encoding method, and the specific process is as follows:

[0018] Collect samples of known ground object categories on the generated low-tide synthetic images and high-tide synthetic images respectively;

[0019] Based on the respective spectral reflectance curves of the low tide and high tide synthetic images and / or the first-order differential spectral reflectance curves based on the spectral reflectance curves, the spectral reflectance mean values ​​and / or the first-order differential spectral reflectance mean values ​​of all ground object categories on the low tide and high tide synthetic images are calculated respectively, and used as the intermediate threshold values ​​T corresponding to the low tide and high tide synthetic images. mid ;

[0020] The minimum value T of the spectral reflectance and / or the first-order differential spectral reflectance of all pixels on the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low tide and high tide synthetic images is obtained respectively. min , maximum value T max , and based on the minimum value T of the same curve min , the maximum value T max And the intermediate threshold T mid Forming interval [T min ,T mid ] and [T mid ,T max ]; that is, the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low-tide synthetic image correspond to a set of intervals [T min ,T mid ] and [T mid ,T max ]; The spectral reflectance curve and / or the first-order differential spectral reflectance curve of the climax synthetic image also correspond to a set of intervals [T min ,T mid ] and [T mid ,T max ];

[0021] Then, the spectral reflectance and / or the first-order differential spectral reflectance of the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low tide and high tide synthetic images respectively falling in the corresponding two intervals are averaged to obtain the corresponding 1 / 4 average threshold value T 1 / 4 With 3 / 4 average threshold T 3 / 4 , and then form the interval corresponding to the low tide and high tide [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ];

[0022] Among them, the interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ] correspond to four character codes respectively;

[0023] Then, according to the spectral reflectance and / or first-order differential spectral reflectance corresponding to each band of the pixel of each ground object category on the low tide and high tide synthetic image and the interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ], and encode according to the band order to form a DNA code chain for each ground feature category corresponding to the low tide and high tide synthetic images;

[0024] Among them, the DNA code chain of each feature category is composed of the codes of all bands;

[0025] Finally, the DNA code chains generated for each type of land feature on the low tide and high tide synthetic images are compared, and the code segments with the greatest difference relative to the mudflats are selected respectively, among which the code segments corresponding to the low tide and high tide are the distinguishing characteristic bands corresponding to the low tide and high tide.

[0026] Further optionally, if encoding is performed based on a spectral reflectance curve and a first-order differential spectral reflectance curve, a reflectance code chain is obtained based on the spectral reflectance curve, and a waveform code chain is obtained based on the first-order differential spectral reflectance curve, and then the reflectance code chain and the waveform code chain of the same land feature category are merged to obtain a DNA code chain.

[0027] Further optionally, the process of generating the low tide synthetic image and the high tide synthetic image by using the quantile synthesis method based on the remote sensing image in step 2 is as follows:

[0028] Step 21: Counting the spectral reflectance corresponding to each pixel in each band of the remote sensing image within the target period, and then generating a histogram corresponding to each pixel in each band;

[0029] Step 22: for each pixel in each band, determine the spectral reflectance whose cumulative proportion reaches the preset low proportion and the preset high proportion respectively, and then calculate the average of the spectral reflectance whose cumulative proportion is less than or equal to the preset low proportion and the average of the spectral reflectance whose cumulative proportion is greater than or equal to the preset high proportion;

[0030] Step 23: Generate a low tide synthetic image using the average spectral reflectance of each pixel in all bands that is less than or equal to a preset low proportion, and generate a high tide synthetic image using the average spectral reflectance of each pixel in all bands that is greater than or equal to a preset high proportion.

[0031] Further optionally, the remote sensing image is a Sentinel-2 image.

[0032] In a second aspect, the present invention provides a system based on the intertidal zone mudflat extraction method, comprising:

[0033] A remote sensing image acquisition module is used to acquire remote sensing images of a target area within a target period;

[0034] A low tide and high tide synthetic image generation module, used for generating a low tide synthetic image and a high tide synthetic image by using a quantile synthesis method based on the remote sensing image;

[0035] A tidal flat index calculation module is used to calculate the tidal flat index corresponding to each pixel position using low tide and high tide synthetic images;

[0036] The intertidal zone mudflat extraction module is used to perform object-oriented multi-scale segmentation using the calculation results of the tidal flat index to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal zone mudflat.

[0037] In a third aspect, the present invention provides an electronic terminal, comprising:

[0038] one or more processors;

[0039] and a memory storing one or more computer programs;

[0040] The processor invokes the computer program to execute:

[0041] The invention discloses a method for extracting intertidal mudflats based on high and low tide synthetic images.

[0042] In a fourth aspect, the present invention provides a readable storage medium storing a computer program, wherein the computer program is called by a processor to execute:

[0043] The invention discloses a method for extracting intertidal mudflats based on high and low tide synthetic images.

[0044] Beneficial Effects

[0045] 1. In view of the problem that intertidal mudflats are affected by tidal fluctuations, which makes it difficult to extract mudflats using remote sensing technology, the present invention provides an intertidal mudflat extraction method based on tidal flat index, which makes full use of remote sensing images within the target period to synthesize low tide synthetic images and high tide synthetic images, and at the same time incorporates the characteristics of the low tide and high tide synthetic images into the calculation of the tidal flat index, so that the calculated tidal flat index has both low tide and high tide characteristics, effectively overcomes the influence of tidal periodic inundation, fills the gap in mudflat extraction technology considering tidal periodic inundation, thereby effectively improving the accuracy of mudflat extraction, and the technical solution of the present invention is easy to implement, which supplements and improves the existing mudflat extraction method, and can quickly, simply and accurately extract the distribution of intertidal mudflats.

[0046] 2. In a further preferred embodiment of the present invention, based on the tidal fluctuation characteristics in the field, the low tide and high tide characteristics are integrated to propose a new calculation formula for the tidal flat index. The tidal flat index calculated by the formula can more effectively and accurately distinguish between tidal flats and other landforms, thereby improving the final accuracy of tidal flat extraction. Among them, the tidal flat index in the preferred embodiment of the present invention enhances the difference in spectral reflectance between tidal flats and other background landforms through square operation, and suppresses the spectral reflectance of background landforms such as water bodies through product operation, thereby highlighting the spectral reflectance characteristics of tidal flats and increasing the significant difference between tidal flats and other landforms.

[0047] 3. In a further preferred embodiment of the present invention, in order to obtain distinguishing characteristic bands for identifying tidal flats and other landform categories, the DNA coding theory is creatively applied to this field for the first time, and an improved DNA coding method suitable for this field is proposed, which effectively improves the selection accuracy of distinguishing characteristic bands, thereby ultimately improving the reliability and accuracy of the tidal flat index, and laying the foundation for the accurate extraction of intertidal tidal flats. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 These are Sentinel-2 images at high and low tide levels, where (a) is a high tide composite image and (b) is a low tide composite image;

[0049] Figure 2The mean spectral reflectance curves of aquaculture ponds, background water, buildings, terrestrial vegetation, and mudflats in Sentinel-2 images at high and low tide levels;

[0050] Figure 3 It is a schematic diagram of the calculation results of the intertidal tidal flat index (TFRI) in the study area;

[0051] Figure 4 It is a box plot generated by the statistical calculation results of the tidal flat index (TFRI) of different landforms;

[0052] Figure 5 It is the final identification result of the intertidal mudflats in the study area;

[0053] Figure 6 It is a flow chart of a method for extracting intertidal mudflats based on high and low tide synthetic images provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The purpose of the present invention is to improve the accuracy of tidal flat extraction and overcome the technical obstacles to tidal flat extraction caused by tidal periodic inundation, thereby proposing a method for extracting intertidal tidal flats based on high and low tide synthetic images, filling the gap in tidal flat extraction technology considering tidal periodic inundation. The present invention will be further explained in conjunction with embodiments below.

[0055] Embodiment 1:

[0056] This embodiment provides a method for extracting intertidal mudflats based on a tidal flat index, which comprises the following steps:

[0057] Step 1: Obtain remote sensing images of the target area within the target period.

[0058] Among them, this embodiment sets the target period to one year, and the remote sensing image is a Sentinel-2 image. All Sentinel-2 images in a year are selected to form a Sentinel-2 image set, and then each Sentinel-2 image in the Sentinel-2 image set is preprocessed. The preprocessing technology includes radiation calibration, atmospheric correction and orthorectification. Among them, since preprocessing is a conventional technology, it is not described in detail, and different preprocessing methods can be selected according to actual needs and accuracy requirements in different embodiments, and the present invention does not impose specific constraints on this. It should also be understood that in other feasible embodiments, the target period can be adaptively adjusted according to actual needs, and other remote sensing images suitable for the technical ideas of the present invention also belong to the protection scope of the present invention.

[0059] Step 2: Based on the remote sensing image, a quantile synthesis method is used to generate a low tide synthetic image and a high tide synthetic image. Among them, this embodiment selects to use the quantile synthesis method to generate low tide and high tide Sentinel-2 synthetic images. The specific process is as follows:

[0060] Step 21: Count the spectral reflectance of each pixel in each band in the Sentinel-2 image set, and then generate a histogram corresponding to each pixel in each band.

[0061] The Sentinel-2 image includes 12 bands, and each pixel corresponds to a spectral reflectance in each band. In this embodiment, the spectral reflectance of each pixel (same pixel position) in each band of the Sentinel-2 image set is counted, and the values ​​are arranged from small to large, and a histogram is generated for the arranged values.

[0062] Step 22: For the histogram corresponding to each pixel in each band, determine the spectral reflectance whose cumulative proportion reaches the preset low proportion and the preset high proportion respectively, and then calculate the average spectral reflectance whose cumulative proportion is less than or equal to the preset low proportion and the average spectral reflectance whose cumulative proportion is greater than or equal to the preset high proportion.

[0063] Step 23: Generate a low tide synthetic image using the average spectral reflectance of each pixel in all bands that is less than or equal to a preset low proportion, and generate a high tide synthetic image using the average spectral reflectance of each pixel in all bands that is greater than or equal to a preset high proportion.

[0064] Among them, in this embodiment, the preset low proportion is set to 10%, and the preset high proportion is set to 90%; thus, for the histogram corresponding to each pixel in each band, this embodiment determines and marks the location of the spectral reflectance with a cumulative proportion of 10% and 90%, and then calculates the average of the spectral reflectances less than or equal to 10% and greater than or equal to 90% in the histogram. That is, the above calculation is repeated for the 12 bands corresponding to each pixel, and finally the low tide Sentinel-2 synthetic image is synthesized using the average of the spectral reflectances obtained by each pixel in the 12 bands that are less than or equal to 10%; and the high tide Sentinel-2 synthetic image is synthesized using the average of the spectral reflectances obtained by each pixel in the 12 bands that are greater than or equal to 90%.

[0065] It should be understood that in other feasible embodiments, the preset low ratio and the preset high ratio can be adaptively adjusted according to application requirements and accuracy requirements.

[0066] Step 3: Calculate the tidal flat index corresponding to each pixel using the low tide and high tide synthetic images.

[0067] The preferred calculation formula of the tidal flat index in this embodiment is as follows:

[0068]

[0069] Where TFRI is the tidal flat index, is the spectral reflectance of the i-th band on the low-tide synthetic image, and is the spectral reflectance of the j-th band and the k-th band on the high-tide synthetic image, respectively.

[0070] Among them, the tidal flat index of the Sentinel-2 image of the present invention enhances the difference in spectral reflectance between the tidal flat and other background objects through a square operation, and suppresses the spectral reflectance of background objects such as water bodies through a product operation, thereby highlighting the spectral reflectance characteristics of the tidal flat.

[0071] In this embodiment, the i-th band is the 8th band, the j-th band is the 6th band, and the k-th band is the 7th band. Bands 6 (0.740), 7 (0.783), and 8 (0.842) are separable. The near-infrared band of band 8 (0.842) is sensitive to water bodies. Bands 6 (0.740) and 7 (0.783) are red-edge bands, which can extract tidal flats according to the difference in spectral reflectance. According to the selected reflectance characteristic bands, the tidal flat index TFRI (intertidal flat extraction index) of the above-mentioned Sentinel-2 image is constructed. In this embodiment, considering that band 9 is the turning point of many land feature categories (the turning point of the upward or downward trend), if the turning point band 9 is selected, it will bring great uncertainty and the effect is not as good as the above-mentioned bands. Therefore, the above-mentioned setting of this embodiment is the best value. In other feasible embodiments, it can be calculated according to the following method. Among them, in order to obtain more accurate distinguishing characteristic bands, this embodiment preferably adopts DNA encoding method for extraction, and the specific implementation process will be highlighted below. In other feasible embodiments, other existing methods may also be used to obtain the information. In short, the identification criterion for distinguishing characteristic bands is the band with the most significant difference between the data of the tidal flat and other land object categories.

[0072] Step 4: Use the calculation result of the tidal flat index to perform object-oriented multi-scale segmentation to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal flats in the target area.

[0073] Among them, when the tidal flat index of the present invention is used for segmentation, the tidal flat and other land object categories can be effectively separated, so as to obtain more accurate tidal flat patches. The technical means for performing object-oriented multi-scale segmentation based on the tidal flat extraction index is a prior art, so it is not specifically stated. In this embodiment, the classification model is set as a remote sensing classification method such as an expert decision tree, a neural network, a support vector machine, and a nearest algorithm, and a classification model trained by a remote sensing classification method using training samples of different land object categories. Among them, the input of the classification model is the characteristics of the plaque, and the output result is the type of the plaque; or the input of the classification model is the characteristics of the plaque, and the output result is the classification result of whether the plaque is an intertidal tidal flat, and the selected classification algorithm and input characteristics are referenced to the prior art, so it is not specifically stated. In addition, the training samples of the classification model are preferably samples generated by the same type of remote sensing images, and artificial marking or other high-precision marking algorithms can be used to determine the tidal flat patches and other non-tidal flat patches on the samples. When there are different requirements (lower) for the model accuracy, the source of the training samples of the classification model can be adaptively adjusted.

[0074] This embodiment can effectively improve the accuracy of tidal flat extraction through the above method, and obtain more accurate distribution results of intertidal tidal flats.

[0075] Regarding the selection of distinguishing characteristic bands, it is preferred to use DNA encoding method for extraction, and the specific implementation process is as follows:

[0076] S1: Collect samples of known ground object categories on the generated low tide synthetic images and high tide synthetic images respectively.

[0077] Among them, the ground object category samples of the same type of remote sensing images with the same target cycle length are preferred. In this embodiment, the ground object categories include: breeding ponds, background water bodies, buildings, terrestrial vegetation, and tidal flats.

[0078] S2: Based on the respective spectral reflectance curves of the low tide and high tide synthetic images and / or the first-order differential spectral reflectance curves based on the spectral reflectance curves, the spectral reflectance mean values ​​and / or the first-order differential spectral reflectance mean values ​​of all ground object categories on the low tide and high tide synthetic images are calculated respectively, and used as the intermediate threshold T corresponding to the low tide and high tide synthetic images. mid .

[0079] In this embodiment, it is preferred to select typical sample points in each land feature category to form a mean spectral reflectance curve of the above land feature category on the low tide synthetic image and a first-order differential spectral reflectance curve based on the mean spectral reflectance curve, and perform subsequent processing on the mean spectral reflectance curve and the first-order differential spectral reflectance curve according to the following method, that is, to obtain two intermediate thresholds T of the spectral reflectance curve and the first-order differential spectral reflectance curve for the low tide synthetic image. mid; For the high tide synthetic image, the two intermediate thresholds T of the spectral reflectance curve and the first-order differential spectral reflectance curve are obtained mid .

[0080] In other feasible embodiments, selecting typical sample points or selecting all image points of each land object category to calculate the mean spectral reflectance curve of each land object category or calculating the first-order differential spectral reflectance curve of each land object category are all technical means protected by the present invention. In other feasible embodiments, instead of selecting the mean value calculation, the spectral emissivity curve and the first-order differential spectral reflectance curve of the typical sample points in each land object category are generated in units of image points, and each curve is subsequently processed according to the following method, and the final result is finally obtained comprehensively. It is feasible, but it will greatly increase the amount of calculated data.

[0081] In other feasible embodiments, it is also possible to use only the mean spectral reflectance curve of the above-mentioned land feature categories on the low tide synthetic image or the first-order differential spectral reflectance curve based on the mean spectral reflectance curve to execute subsequent processes to identify and distinguish characteristic bands, and its effect is weaker than the combination of the two types of curves in this embodiment.

[0082] S3: Obtain the minimum value T of the spectral reflectance and / or the first-order differential spectral reflectance of all pixels on the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low tide and high tide synthetic images respectively. min , maximum value T max , and based on the minimum value T min , the maximum value T max And the intermediate threshold T mid Forming interval [T min ,T mid ] and [T mid ,T max ].

[0083] In this embodiment, a set of intervals [T min ,T mid ] and [T mid ,T max ]; For the mean spectral reflectance curve and the first-order differential spectral reflectance curve of the high tide, a set of intervals [T min ,T mid ] and [T mid ,T max ].

[0084] In other feasible embodiments, the corresponding processing is performed according to the curve generated in the previous step S2 to obtain a respective set of intervals [T min ,T mid ] and [Tmid ,T max ].

[0085] S4: Then, the spectral reflectance and / or the first-order differential spectral reflectance of the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low tide and high tide synthetic images respectively falling in the corresponding two intervals are averaged to obtain the corresponding 1 / 4 average threshold value T 1 / 4 With 3 / 4 average threshold T 3 / 4 , and then form the interval corresponding to the low tide and high tide [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ];

[0086] Similarly, in this embodiment, for the mean spectral reflectance curve and the first-order differential spectral reflectance curve of low tide, a set of intervals [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ]; For the mean spectral reflectance curve and the first-order differential spectral reflectance curve of the high tide, a set of intervals [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ].

[0087] In other feasible embodiments, the corresponding processing is performed according to the curve generated in the previous step S2 to obtain a respective set of intervals [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ].

[0088] S5: Based on the spectral reflectance and / or first-order differential spectral reflectance corresponding to each ground feature category pixel in each band on the low tide and high tide synthetic image and the interval [T min ,T 1 / 4 ),[T1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ] relationship, and encode in band order to form a DNA code chain for each land feature category corresponding to the low tide and high tide synthetic images; wherein the DNA code chain for each land feature category is composed of the encodings of all bands.

[0089] Among them, the interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ] correspond to four character codes, for example, four intervals [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ], the four intervals correspond to the four encoding methods of A, T, C, and G respectively.

[0090] In this embodiment, the data of the mean spectral reflectance curve and the first-order differential spectral reflectance curve at low tide and the corresponding interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ], and encode them in band order to form the reflectance code chain and waveform code chain of each feature category corresponding to low tide, and then merge the reflectance code chain and waveform code chain of the same feature category to obtain the DNA code chain of low tide of the feature category; for the data of the mean spectral reflectance curve and the first-order differential spectral reflectance curve of high tide and the corresponding interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ] are encoded in band order to form a reflectance code chain and a waveform code chain for each land feature category corresponding to the climax, and then the reflectance code chain and waveform code chain of the same land feature category are merged to obtain the DNA code chain of the climax of the land feature category.

[0091] In other feasible embodiments, if there are multiple mean spectral reflectance curves and / or first-order differential spectral reflectance curves corresponding to each land feature category in the low tide and high tide synthetic images, then multiple low tide DNA code chains and high tide DNA code chains are obtained accordingly.

[0092] S6: Compare the DNA code chains generated for each type of land feature on the low tide and high tide synthetic images, and select the code segments with the greatest difference relative to the mudflats, wherein the code segments corresponding to the low tide and high tide obtained are the distinguishing characteristic bands corresponding to the low tide and high tide.

[0093] In this embodiment, based on the principle of maximum intra-class similarity and maximum inter-class difference, the difference in DNA code chains between tidal flats and other landform categories is analyzed, and the band with the greatest difference is selected.

[0094] It should be understood that this embodiment applies the DNA encoding method to this field for the first time to determine the distinguishing characteristic bands. Compared with conventional extraction methods, such as the envelope method, the characteristic parameter method, and the differential method, the conventional methods all form a spectral curve with certain characteristics, visually read out the characteristic bands, and then extract the bands, which are subjective and arbitrary. However, the DNA encoding method can be used to quantify and extract the characteristic bands, which can not only judge the trend of the entire waveform as a whole through the DNA encoding chain, but also have the sensitive ability to perceive the changes caused by subtle differences in the bands, which is helpful for the construction of the tidal flat index based on the extracted bands in the later stage, making the classification results of the tidal flats more reliable.

[0095] It should be noted that after obtaining the distinguishing characteristic bands, the Band Math tool in the ENVI software can be used to calculate the intertidal zone mudflat index corresponding to the sample, such as Figure 3 As shown; then quantitatively count the calculation results of the tidal flat index of the sample points in the aquaculture pond (pond), background water body, building, terrestrial vegetation, and tidal flat to generate the following Figure 4 The box plot shown is used to determine the separability of the tidal flats from other landforms. If the separability meets the requirements, it is considered that the obtained distinguishing characteristic bands meet the requirements.

[0096] For example, the present invention selects the Beibu Gulf waters of Zhanjiang City, Guangdong Province as the study area, selects all Sentinel-2 images of the study area in 2021, a total of 249 scenes, to form a Sentinel-2 image set. After quantile synthesis method processing, the following is obtained Figure 1 Second, typical sample points were selected in the aquaculture ponds, background water bodies, buildings, terrestrial vegetation, and mudflats to analyze the low tide and high tide Sentinel-2 synthetic images of the above-mentioned objects. Figure 2 The mean spectral reflectance curve of . Finally, we get Figure 5 Distribution map of intertidal flats shown.

[0097] Embodiment 2:

[0098] This embodiment provides a system based on the intertidal zone mudflat extraction method, which includes: a remote sensing image acquisition module, a low tide and high tide synthetic image generation module, a distinguishing characteristic band acquisition module, a tidal flat index calculation module, and an intertidal zone mudflat extraction module.

[0099] Among them, the remote sensing image acquisition module is used to obtain the remote sensing image of the target area within the target period.

[0100] The low tide and high tide synthetic image generation module is used to generate low tide synthetic images and high tide synthetic images based on the remote sensing image by using the quantile synthesis method. The technical details of the quantile synthesis method can refer to the corresponding content of the above method.

[0101] The distinguishing characteristic band acquisition module is used to obtain the distinguishing characteristic bands corresponding to the low tide synthetic image and the high tide synthetic image. Among them, the DNA encoding method can be used, and other existing methods can also be used. For details, please refer to the above content.

[0102] The tidal flat index calculation module is used to calculate the tidal flat index corresponding to each pixel position using low tide and high tide synthetic images. Please see the previous article for the specific formula.

[0103] The intertidal flat extraction module is used to perform object-oriented multi-scale segmentation using the calculation results of the tidal flat index to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal flats. Among them, the classification model is a direct reference to an existing classification model or a retrained classification model. If it is retrained, the corresponding system also includes a classification model construction module for constructing a classification module.

[0104] Please refer to the above method for the specific implementation process of each module, which will not be repeated here. It should be understood that the division of the above functional modules is only a division of logical functions, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. At the same time, the above integrated units can be implemented in the form of hardware or software functional units.

[0105] Embodiment 3:

[0106] This embodiment provides an electronic terminal, which includes one or more processors and a memory storing one or more computer programs, wherein the processor calls the computer program to execute: a method for extracting intertidal flats based on tidal flat index. Specifically, the following steps are performed:

[0107] Step 1: Acquire remote sensing images of the target area within the target period;

[0108] Step 2: generating a low tide synthetic image and a high tide synthetic image by using a quantile synthesis method based on the remote sensing image;

[0109] Step 3: Calculate the tidal flat index corresponding to each pixel using the low tide and high tide synthetic images;

[0110] Step 4: Use the calculation result of the tidal flat index to perform object-oriented multi-scale segmentation to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal flats in the target area.

[0111] In some implementations, when the processor calls the computer program to implement step 2, steps 21-23 are also executed.

[0112] The memory may include a high-speed RAM memory, and may also include a non-volatile memory, such as at least one disk memory.

[0113] If the memory and the processor are implemented independently, the memory, the processor and the communication interface can be connected to each other through a bus and communicate with each other. The bus can be an industrial standard architecture bus, an external device interconnection bus or an extended industrial standard architecture bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0114] Optionally, in a specific implementation, if the memory and the processor are integrated on a chip, the memory and the processor can communicate with each other through an internal interface.

[0115] For the specific implementation process of each step, please refer to the description of the above method.

[0116] It should be understood that in the embodiments of the present invention, the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the type of device.

[0117] Embodiment 4:

[0118] This embodiment provides a readable storage medium storing a computer program, wherein the computer program is called by a processor to execute: steps of a method for extracting intertidal mudflats based on a tidal flat index. Specifically, the following steps are performed:

[0119] Step 1: Acquire remote sensing images of the target area within the target period;

[0120] Step 2: generating a low tide synthetic image and a high tide synthetic image by using a quantile synthesis method based on the remote sensing image;

[0121] Step 3: Calculate the tidal flat index corresponding to each pixel using the low tide and high tide synthetic images;

[0122] Step 4: Use the calculation result of the tidal flat index to perform object-oriented multi-scale segmentation to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal flats in the target area.

[0123] In some implementations, when the processor calls the computer program to implement step 2, steps 21-23 are also executed.

[0124] For the specific implementation process of each step, please refer to the description of the above method.

[0125] The readable storage medium is a computer-readable storage medium, which may be an internal storage unit of the controller described in any of the foregoing embodiments, such as a hard disk or memory of the controller. The readable storage medium may also be an external storage device of the controller, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped on the controller. Further, the readable storage medium may also include both an internal storage unit of the controller and an external storage device. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0126] Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned readable storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0127] It should be emphasized that the examples described in the present invention are illustrative rather than restrictive, and therefore the present invention is not limited to the examples described in the specific embodiments. Any other embodiments derived by those skilled in the art based on the technical solution of the present invention that do not depart from the purpose and scope of the present invention, whether modified or replaced, also fall within the scope of protection of the present invention.

Claims

1. A method for extracting intertidal flats based on tidal flat index, characterized in that: The following steps are involved: Step 1: Acquire remote sensing images of the target area within the target period; Step 2: generating a low tide synthetic image and a high tide synthetic image by using a quantile synthesis method based on the remote sensing image; Step 3: Calculate the tidal flat index corresponding to each pixel position using the low tide and high tide synthetic images; Step 4: Use the calculation result of the tidal flat index to perform object-oriented multi-scale segmentation to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal flats in the target area; The calculation formula of the tidal flat index is as follows: Where TFRI is the tidal flat index, Bandi low is the spectral reflectance of the ith band at a certain pixel position in the low tide synthetic image, Bandj high and Bandk high are the spectral reflectances of the j-th band and the k-th band at the same pixel position in the high tide synthetic image, respectively; Among them, the i-th band is a distinguishing characteristic band obtained based on the low-tide synthetic image for identifying the tidal flats and other landforms; the j-th and k-th bands are based on the high-tide synthetic image for identifying the tidal flats and other landforms; The i-th band, the j-th band, and the k-th band are all distinguishing characteristic bands extracted by the DNA encoding method, and the specific process is as follows: Collect samples of known ground object categories on the generated low-tide synthetic images and high-tide synthetic images respectively; Based on the respective spectral reflectance curves of the low tide and high tide synthetic images and / or the first-order differential spectral reflectance curves based on the spectral reflectance curves, the spectral reflectance mean values ​​and / or the first-order differential spectral reflectance mean values ​​of all ground object categories on the low tide and high tide synthetic images are calculated respectively, and used as the intermediate threshold values ​​T corresponding to the low tide and high tide synthetic images. mid ; The minimum value T of the spectral reflectance and / or the first-order differential spectral reflectance of all pixels on the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low tide and high tide synthetic images is obtained respectively. min , maximum value T max , and based on the minimum value T of the same curve min , the maximum value T max And the intermediate threshold T mid Form a set of intervals [T min ,T mid ] and [T mid ,T max ]; Then, the spectral reflectance curve and / or the first-order differential spectral reflectance curve of the low tide and high tide synthetic images fall in the corresponding interval [T min ,T mid ] and [T mid ,T max ] and / or the first-order differential spectral reflectance are averaged to obtain the corresponding 1 / 4 average threshold value T 1 / 4 With 3 / 4 average threshold T 3 / 4 , and then form the intervals corresponding to the low tide and high tide [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ]; Among them, the interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ] correspond to four character codes respectively; Then, according to the spectral reflectance and / or first-order differential spectral reflectance corresponding to each band of the pixel of each ground object category on the low tide and high tide synthetic image and the interval [T min ,T 1 / 4 ),[T 1 / 4 ,T mid ),[T mid ,T 3 / 4 ) and [T 3 / 4 ,T max ], and encode according to the band order to form a DNA code chain for each ground feature category corresponding to the low tide and high tide synthetic images; Among them, the DNA code chain of each feature category is composed of the codes of all bands; Finally, the DNA code chains generated for each type of land feature on the low tide and high tide synthetic images are compared, and the code segments with the greatest difference relative to the mudflats are selected respectively, among which the code segments corresponding to the low tide and high tide are the distinguishing characteristic bands corresponding to the low tide and high tide.

2. The intertidal zone mudflat extraction method according to claim 1, characterized in that: The i-th band is the 8th band, the j-th band is the 6th band, and the k-th band is the 7th band.

3. The intertidal zone mudflat extraction method according to claim 1, characterized in that: If encoding is performed based on a spectral reflectance curve and a first-order differential spectral reflectance curve, a reflectance code chain is obtained based on the spectral reflectance curve, and a waveform code chain is obtained based on the first-order differential spectral reflectance curve. The reflectance code chain and the waveform code chain of the same land feature category are then merged to obtain a DNA code chain.

4. The intertidal zone mudflat extraction method according to claim 1, characterized in that: The process of generating a low tide synthetic image and a high tide synthetic image by using the quantile synthesis method based on the remote sensing image in step 2 is as follows: Step 21: Counting the spectral reflectance corresponding to each pixel in each band of the remote sensing image within the target period, and then generating a histogram corresponding to each pixel in each band; Step 22: for each pixel in each band, determine the spectral reflectance whose cumulative proportion reaches the preset low proportion and the preset high proportion respectively, and then calculate the average of the spectral reflectance whose cumulative proportion is less than or equal to the preset low proportion and the average of the spectral reflectance whose cumulative proportion is greater than or equal to the preset high proportion; Step 23: Generate a low tide synthetic image using the average spectral reflectance of each pixel in all bands that is less than or equal to a preset low proportion, and generate a high tide synthetic image using the average spectral reflectance of each pixel in all bands that is greater than or equal to a preset high proportion.

5. The intertidal zone mudflat extraction method according to claim 1, characterized in that: The remote sensing image is a Sentinel-2 image.

6. A system based on the intertidal zone mudflat extraction method according to any one of claims 1 to 4, characterized in that: include: A remote sensing image acquisition module is used to acquire remote sensing images of a target area within a target period; A low tide and high tide synthetic image generation module, used for generating a low tide synthetic image and a high tide synthetic image by using a quantile synthesis method based on the remote sensing image; A tidal flat index calculation module is used to calculate the tidal flat index corresponding to each pixel position using low tide and high tide synthetic images; The intertidal zone mudflat extraction module is used to perform object-oriented multi-scale segmentation using the calculation results of the tidal flat index to obtain patch objects, and then use the classification model to classify the patch objects to obtain the distribution results of the intertidal zone mudflat.

7. An electronic terminal, characterized in that: include: one or more processors; and a memory storing one or more computer programs; The processor invokes the computer program to execute: The steps of the intertidal flat extraction method according to any one of claims 1 to 5.

8. A readable storage medium, characterized in that: A computer program is stored, which is called by a processor to execute: The steps of the intertidal flat extraction method according to any one of claims 1 to 5.

Images