A method for making an initial field of distribution area and drift prediction of green tide in the Yellow Sea based on a distance vertex angle rule

Abstract

A kind of making method of initial field of distribution area and drift prediction of yellow sea green tide based on distance vertex angle rule, selects vertex as prediction initial field by hand in distribution range boundary, and establishes vertex extraction rule based on distance and angle constraint using this initial field analysis, and the method obtains initial field can serve business or emergency monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of model design and prediction, in particular, the present application relates to a method for making an initial field of a distribution area of green tide in the Yellow Sea and drift prediction based on a distance top corner rule. BACKGROUND

[0002] The outbreak of green tide in the Yellow Sea causes different degrees of harm to marine fishery breeding, transportation, tourism, water sports, etc. in the south coast of Shandong, so the marine management departments in the disaster area will intercept and salvage the green algae. The distribution area information and drift trend of green tide are the main reference basis for the layout of green algae interception nets and the scheduling of salvage vessels.

[0003] CN201911249340.7 discloses a green tide biomass prediction method, device, equipment and medium. The method comprises: determining an initially constructed green tide biomass estimation model; wherein the green tide biomass estimation model comprises at least one undetermined parameter; determining the value of the at least one undetermined parameter included in the initially constructed green tide biomass estimation model according to reference distribution data of green tide biomass in a target area to obtain a green tide biomass estimation model after parameter determination, wherein the reference distribution data is determined according to satellite remote sensing images.

[0004] CN202210253275.0 discloses a method for predicting the medium and long-term trend of green tide in the Yellow Sea, comprising the following steps: a. determining the research area of green tide in the Yellow Sea (33-37 °N, 119-123 °E), and dividing it into two key areas of green tide generation and development with 35 °N as the boundary, and determining the main factors affecting the growth and drift of green tide as meteorological and marine factors; b. obtaining green tide multi-source monitoring data and meteorological and marine element observation data in the key area; c. analyzing the pre-period meteorological and marine influence factors of three indexes of green tide satellite discovery time, green tide main drift direction and green tide maximum distribution area, and establishing a prediction model respectively; d. obtaining the meteorological element values and marine element values required for the prediction of the pre-period of the green tide occurrence in the required prediction year, and according to the prediction model established in step c, the medium and long-term trend of the satellite discovery time, the main drift direction and the maximum distribution area of the green tide in the Yellow Sea in the current year are predicted, and the prediction result is obtained.

[0005] Satellite remote sensing has the advantages of instantaneity and wide range, and is the main data source for monitoring green tide in the Yellow Sea. It is also the only means to obtain complete and comprehensive green tide information. In the response to green tide disasters in the Yellow Sea, satellite images are used to monitor green tide coverage information, green tide coverage information is used to make distribution areas, and the vertices of the distribution areas are used as initial fields to predict the drift of green tide distribution.

[0006] The distribution of green tide is defined as the outer envelope surface of the coverage area of Enteromorpha prolifera. There are two existing methods for obtaining the distribution area: the first method is manual drawing, in which a satellite remote sensing monitor draws the distribution surface along the outer edge of the coverage based on the coverage information of green tide. The advantage of manual drawing is that the number of vertices of the distribution polygon is moderate, which can be directly used as an initial field, and the time-consuming of drift prediction calculation is short. The disadvantage is that it is highly subjective and has no uniform standard, and different monitors draw different distribution surfaces. The second method is the buffer zone method, in which a buffer tool in the ArcGIS toolbox is used to generate the envelope surface by buffering the coverage area by a certain distance. The number of vertices of the distribution polygon generated by this method is huge, often from tens of thousands to hundreds of thousands, and needs to be thinned. The existing thinning method of ArcGIS is to extract vertices at equal intervals (i.e. the equal interval method). When the number of vertices after extraction is comparable to that of manual drawing, the distribution polygon is greatly deformed, and some coverage points of green tide are missed, which need to be manually corrected. Therefore, the buffer zone method is used to extract the vertices of the distribution, and the advantage of this method is that it can automatically form a unified and standardized distribution area, and the disadvantage is that the number of vertices of the distribution area is huge, and there is no standardized and scientific method for extraction.

[0007] Considering that the existing two methods of distribution area and initial field have advantages and disadvantages, the present application takes into account the standardization and automation of the two aspects, and combines the advantages of the ArcGIS buffer zone analysis and manual selection of vertices to establish an automatic vertex thinning method. The vertex thinning method can sufficiently maintain the shape of the distribution after thinning, and the number of points is reduced to be comparable to that of manual selection, greatly reducing the calculation amount of drift prediction and improving the operation efficiency. SUMMARY

[0008] In view of the problems in the prior art, the present application provides a method for making a distribution area of green tide in the Yellow Sea and an initial field for drift prediction thereof based on a distance vertex angle rule.

[0009] The method of the present application has three advantages in obtaining the initial field: first, the number of vertices is small, which reduces the prediction time; second, the small number of vertices can sufficiently maintain the shape of the distribution surface; and third, the method can automatically obtain a standardized result, reducing manpower, saving time and improving the emergency level.

[0010] The method for making a distribution area of green tide in the Yellow Sea and an initial field for drift prediction thereof of the present application can serve the business or emergency monitoring of green tide in the Yellow Sea.

[0011] A method for making a distribution area of green tide in the Yellow Sea and an initial field for drift prediction thereof based on a distance vertex angle rule, comprising:

[0012] Step 1: Calculate the normalized vegetation index NDVI using satellite images, and then extract the green tide coverage points by threshold method,

[0013] In the method, the green tide disaster information is extracted by using a standard false color image B4B3B2 (R-nir G-r B-g) and a threshold segmentation of a normalized difference vegetation index (NDVI).

[0014] In the second step, the distribution area is generated by using the buffer space analysis method with the covering points, and a buffer area is generated for the green tide covering points with a radius of 1-5 km, and the buffer areas are combined as the green tide distribution area.

[0015] In the third step, the vertex sparseness rule based on the distance-angle constraint (DACR) is used to sparsify the distribution boundary, and the vertex sparseness rule is determined according to the selected vertex and the distance and vertex angle range as follows:

[0016] d∈[2.5km,6km]andα∈[120°-180°]

[0017] wherein α is the vertex angle and d is the distance between the adjacent vertices.

[0018] In the fourth step, the x and y longitude and latitude coordinates are added to the sparsified vertices to form the initial field of the drift prediction.

[0019] Further, in the first step, the images of the initial stage, development stage, outbreak stage and decline stage of the green tide are collected by using the satellite.

[0020] Further, in the first step, the normalized difference vegetation index is based on the unique spectral characteristics of the green algae red band and near-infrared band, and the formula is: NDVI=(R nir -R red ) / (R nir +R red ), wherein R nir , R red are the near-infrared and red band reflectivities.

[0021] Further, in the first step, the NDVI threshold method is used to extract the green tide information in the green tide area, the histogram analysis of the green tide area image is carried out, the NDVI threshold is determined, and the threshold is floating around 0. Under normal circumstances, the NDVI threshold is 0, and under the condition of thin cloud and fog interference, the NDVI threshold is fine-tuned.

[0022] Further, the area error is used to evaluate the degree of coincidence between the distribution polygon after the vertex sparsification and the original distribution polygon, and the area error is the ratio of the difference part area of the polygon after the sparsification to the area of the original polygon. The difference part area includes the sum of the increased area S i and the reduced area S d . The area error formula is defined as follows:

[0023] R ac = (S i +S d ) / S0 (1)

[0024] S0 is the area before sparsification, S i is the increased area, S d is the reduced area.

[0025] The satellite remote sensing-based Yellow Sea green tide distribution area and its drift prediction initial field production method of the present application has the following advantages in obtaining the initial field:

[0026] First, both standardization and automation are taken into account, and an automatic vertex sparsification method is established by combining the advantages of ArcGIS buffer analysis and manual selection of vertices.

[0027] Second, the number of sparsified vertices is small, reducing the prediction time.

[0028] Third, the small number of sparsified vertices can fully maintain the shape of the distribution surface.

[0029] Fourth, this method can automatically obtain standardized and unified results, reducing manpower, saving time, and improving emergency response level.

[0030] The Yellow Sea green tide distribution area and its drift prediction initial field production method of the present application can serve the Yellow Sea green tide business or emergency monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0031] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0032] Figure 1 is a distribution area vertex selection schematic diagram A and schematic diagram B, respectively for flat area and large curvature area

[0033] Figure 2 is a pentagon sparsification to triangle schematic diagram

[0034] Figure 3 is a green tide coverage and distribution schematic diagram on May 17, 2021

[0035] Figure 4 A is a manual extraction result schematic diagram, Figure 4 B is a manual extraction result vertex distance and vertex angle distribution diagram

[0036] Figure 5 is a sparsification model application process

[0037] Figure 6is the sparse result after the same point interval vertex on August 5, 2021, and the details of the problem are shown on August 5, and similar on other dates DETAILED DESCRIPTION

[0038] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the application by using terminology for the purpose of description only and is not intended to limit the exemplary embodiments based on the present application.

[0039] Example 1

[0040] A method for making an initial field of a distribution area of green tide in the Yellow Sea and drift prediction based on a distance-angle sparse rule, comprising:

[0041] Step 1, calculate the normalized vegetation index NDVI using satellite images, and then extract the green tide coverage points by threshold method,

[0042] For the data and method section, satellite data is used.

[0043] HY-1D satellite was launched on June 11, 2020, and realized dual-satellite networking with HY-1C satellite, which can realize 3-day 2-time full-coverage imaging monitoring of green tide in the Yellow Sea.

[0044] The Coastal Zone Imager carried by HY-1C / D satellite has a spatial resolution of 50m and a width of 950km, containing four bands of blue (0.42μm-0.50μm), green (0.52μm-0.60μm), red (0.61μm-0.69μm), and near-infrared (0.76μm-0.89μm), which can effectively monitor green tide information.

[0045] Since 2021, it has become one of the main sources of business / emergency monitoring of green tide in the Yellow Sea.

[0046] The present application selects 5 images of the initial, development, outbreak and decline stages of green tide (see Table 1 below) for research.

[0047] Table 1 Image information

[0048]

[0049] For the part of green tide information extraction, since the study area is located in the nearshore sea area, the water environment is complex, in order to as far as possible not to miss the Enteromorpha information and reduce the influence of human factors, let the result be closer to the true value, the standard false color image B 432 (R-nir G-r B-g) combined with normalized vegetation index NDVI (Normalized Difference Vegetation Index) threshold segmentation semi-automatically completes the extraction of Enteromorpha disaster information.

[0050] In the false color image, large-scale algae blooms are shown as red, which has stronger contrast with water than the true color image, and is conducive to visual interpretation of green tide. The normalized vegetation index is based on the unique spectral characteristics of green algae in the red and near-infrared bands, and the formula is: NDVI = (R nir -R red ) / (R nir +R red ), where R nir , R red are the reflectivity of near-infrared and red bands.

[0051] The normalized vegetation index can effectively enhance the green tide information and improve the contrast between green tide and water body, so the combination of the two can effectively identify the green tide area.

[0052] Then the NDVI threshold method is used to extract green tide information in the green tide area.

[0053] The green tide information extraction in the green tide area can effectively remove cloud and fog interference information and improve the accuracy of green tide information.

[0054] The image histogram analysis of the green tide area is carried out to determine the NDVI threshold, which is floating around 0. Usually, the NDVI threshold is 0, and in the case of thin cloud and fog interference, the NDVI threshold will be fine-tuned.

[0055] Step 2, use the coverage point to generate the distribution area by buffer spatial analysis method.

[0056] The green tide distribution area is made from the green tide coverage point file. The Buffer tool in ArcToolboxs of ArcGIS desktop software is used to make a buffer zone with a radius of 3 km for the green tide coverage point. The buffer zones formed by each green tide point are merged as the green tide distribution area.

[0057] Step 3, use the vertex sparsification rule based on distance-angle constraint to sparsify the distribution boundary.

[0058] The number of vertices of the green tide distribution polygon is huge, often tens of thousands to hundreds of thousands. If it is directly used as the initial field of distribution for numerical calculation to predict the drift of green tide, the calculation amount is very large and the time-consuming is long, which cannot meet the demand of emergency monitoring and prediction. It needs to be sparsified. After sparsification, the number of vertices should be as small as possible, and the shape of the original polygon should be effectively maintained.

[0059] Sparsification of green tide distribution polygon vertices.

[0060] In order to make the number of selected vertices as small as possible, the distance between points must be as long as possible. For relatively flat boundaries, the distance between vertices can be far enough to maintain the shape of the boundary, such as Figure 1The boundary line between points A1 and A2 in A is long enough while maintaining the shape of the boundary well. The distances B1 and B2 between the vertices of the zigzag boundary are too large, resulting in large deformation. Also, the polygon formed by selecting vertices cannot include algae-covered points. If point B2′ is added, the distance can be reduced, which can effectively maintain the shape. It should be noted that the red dots in the figure are the vertices of the distributed polygon formed by the buffer analysis, which are displayed as lines because they are very dense.

[0061] Analyzing the example above, we can see that in the vertex selection process, considering only distance cannot simultaneously guarantee that suitable vertices will be selected in both straight and curved regions. The angles between a vertex and its adjacent vertices must also be considered. Therefore, constructing a vertex sparsity model requires considering both the distances between vertices and the angles formed by a vertex and its adjacent vertices.

[0062] To address this, this application manually selects vertices from the distribution polygon generated by buffer analysis, minimizing the number of vertices while effectively preserving the distribution shape. Then, it analyzes the patterns of distances between adjacent vertices and the angles between each inflection point and its adjacent points, ultimately establishing a vertex extraction model for the green tide distribution polygon.

[0063] To evaluate the degree of similarity between the distributed polygons after vertex sparsification and the original distributed polygons, area error can be used for verification. A small area error indicates minimal change in the distributed polygons after sparsification, indicating a good sparsification effect. A large area error indicates significant change in the distributed polygons after vertex sparsification, indicating a poor sparsification effect. The area error is the ratio of the area of ​​the difference between the sparsified polygons and the original polygons to the area of ​​the original polygons. The area of ​​the difference includes the increased area S. i and reduce area S d The sum of the areas. The area error formula is defined as follows:

[0064] R = (S) i +S d ) / S0 (1)

[0065] S0 is the area of ​​the original polygon before sparsification, S i To increase the area, S d To reduce the area.

[0066] by Figure 2 For example, if the original polygon is a pentagon ABCDE, after sparsening the vertices, it becomes a triangle ACD. Then, the area S increases here. i For S △ADE Reduce area S d For S △ABC The area of ​​the original pentagon is S ABCDE Therefore, the formula for calculating the area error R here is (S △ABC +S △ADE ) / S ABCDEIn the extreme case, if only 1 or 2 vertices are left after the sparsification, the polygon is lost, and the area error R is 1.

[0067] In order to evaluate whether the sparsification effectively improves the prediction efficiency, the original initial field and the sparsified initial field are calculated by using the green tide drift prediction model operated by the center, and the calculation time is compared and analyzed.

[0068] The green tide coverage and distribution scene images on May 17, May 22, June 6, July 9, and August 5, 2021 were collected. The green tide coverage area of the image was extracted by using the NDVI threshold method combined with expert experience, and the green points in the figure were obtained. The distribution area of the green tide was obtained by using the buffer spatial analysis method, and the red range in the figure was obtained. The coverage and distribution areas extracted from the five date images are: 5.8 / 12242km 2 , 104 / 42455km 2 , 890 / 45906km 2 , 1210 / 42501km 2 , 57 / 8614km 2 . The number of vertices of the distribution polygon is 56763, 51857, 72543, 101881, and 35510.

[0069] The initial green tide is mainly distributed in the northern Jiangsu radiating sandbar. During the development period, the green tide drifts northward, and the coverage and distribution area increase rapidly. During the outbreak period, the coverage area continues to increase, gradually landing on the south coast of Shandong, and the coverage and distribution area decreases rapidly during the extinction period.

[0070] The vertices are manually selected, and the shape of the original distribution polygon is maintained as much as possible. In the area with rapid change in curvature, the distance is as small as possible, and in the area with slow change in curvature, the distance between the vertices is appropriately increased.

[0071] According to the vertices selected by experts, the rules of the vertex angle and edge length of the sparsified distribution polygon are analyzed, and the distance and angle constraint method (Distance-Angle constraint rule) is proposed.

[0072] First, the distance and angle of the distribution polygon formed by the manually selected vertices that meet the initial field of the drift prediction are calculated. It can be understood that the smaller the distance between the adjacent vertices of the sparsified distribution polygon, the larger the angle corresponding to the vertex, and the shape of the original distribution polygon can be well maintained. The distance and angle double constraint rule can effectively maintain the shape of the area with large curvature.

[0073] The distance and angle distribution range between the vertices selected by experts (see Figure 4B), it can be seen that the vertex angle is between 100°-180° and the distance is between 2.5km-7km. Based on the distance between the sparse distribution vertices and the vertex angle of the experts, combined with the following two facts: 1) the distance between the vertices is approximately small, so the sparse distribution is more consistent with the original distribution; 2) the greater the vertex angle after sparse, the higher the consistency of the sparse distribution with the original distribution, we form the following more strict sparse rules:

[0074] d∈[2.5km,5km]andα∈[120°-180°]

[0075] Wherein, α is the vertex angle, and d is the distance between the vertices and the adjacent vertices.

[0076] For the polygon of the image on May 17, 2021, 287 points are selected as the vertices, and the polygon formed by the vertices has high consistency with the distribution range polygon.

[0077] For model evaluation, interval model evaluation can be added to evaluate the model after translation.

[0078] The vertex sparse method proposed in this application through the vertex distance and vertex angle distribution rules is compared with the equal interval method for extracting vertices, which can be compared and analyzed from the following aspects:

[0079] 1. Area error comparison, that is, comparing the value of formula (1) of the area error of the two;

[0080] 2. Whether the sparse distribution area completely includes the coverage area.

[0081] Overall, the area error of the method proposed in this application is lower than that of the equal vertex interval sparse method, and there is no area loss, which fully meets the requirements of the initial field in time and accuracy.

[0082] In the case of keeping the same vertices after sparse, the area error of the equal vertex interval method is about 2 times higher than that of the method proposed in this application, and the shape cannot be well maintained at the position with large boundary radius, resulting in that the sparse distribution range cannot contain the coverage area, causing the loss of the coverage area, which cannot meet the accuracy requirements of the initial field on the distribution.

[0083] Among them, the area error r of the method provided in this application on August 5 is 1.61%, which is larger than the area error of other dates. This is because the total area is small during the green tide dissipation period, so the area change after vertex sparse accounts for a large proportion, and the area error is slightly larger.

[0084] Table 2 Area change of sparse results of each image

[0085]

[0086] Step 4, the sparse vertex after the drift prediction of the initial field, x, y coordinate formation.

[0087] The model of the present application is applied in 4 steps (see Figure 5 ):

[0088] Step 1, using satellite images to calculate the normalized difference vegetation index (NDVI), and then using threshold method to extract green tide coverage points.

[0089] Step 2, using the coverage points to generate the distribution area by buffer spatial analysis method.

[0090] Step 3, using the vertex sparse rule based on distance-angle constraint to sparse the distribution boundary.

[0091] Step 4, the sparse vertex after the drift prediction of the initial field, x, y coordinate formation.

[0092] The main steps of the process are:

[0093] First, realize the buffer with a radius of 3km, forming the distribution area.

[0094] Second, using the distribution area, the vertex of the distribution boundary is sparse processed to form the initial field of numerical simulation.

[0095] For the application of green tide distribution initial field production, the method of the present application provides sparse results, and the sparse results according to the equal sample interval.

[0096] Image sparse results: list the automatic sparse boundary results, and the equal point sparse results. It can be seen that the sparse vertex has a high degree of coincidence with the distribution boundary, and the shape is well preserved.

[0097] The method of the present application can give the application of interval model, which shows that the interval model can fully meet the extraction of the initial field.

Claims

1. A method for making an initial field of distribution area and drift prediction of Ulva prolifera based on a distance-apex angle rule, comprising: Step 1: calculating a normalized vegetation index NDVI by using satellite images, and then extracting Ulva prolifera coverage points by using a threshold method; wherein the Ulva prolifera disaster information is extracted by using a standard false color image B432 in combination with threshold segmentation of the normalized vegetation index NDVI; Normalized difference vegetation index (NDVI) is based on the unique spectral characteristics of green algae in red and near-infrared bands, and the formula is: NDVI=(R nir -R red ) / (R nir +R red ), wherein R nir , R red is the reflectivity of near-infrared and red bands; the Ulva prolifera information is extracted by using the NDVI threshold method for the Ulva prolifera area, histogram analysis of the Ulva prolifera area image is carried out, the NDVI threshold is determined, and the threshold is floating around 0; usually, the NDVI threshold is 0, and in the case of thin cloud and fog interference, the NDVI threshold is fine-tuned; Step 2: generating the distribution area by using the coverage points and a buffer space analysis method, and making a buffer area with a radius of 1-5 km for the Ulva prolifera coverage points, and combining the buffer area as the Ulva prolifera distribution area; Step 3: sparsifying the distribution boundary by using a vertex sparsification rule based on distance-angle constraints; according to the selected vertex, the vertex sparsification rule is determined according to the distance and the vertex angle range as follows: d∈[2.5km,6km]andα∈[120°-180°] wherein α is the vertex angle, and d is the distance between the vertex and the adjacent vertex; Step 4: attaching x, y longitude and latitude coordinates to the sparsified vertex to form the initial field of drift prediction.

2. The method of manufacturing according to claim 1, wherein, In Step 1, the images of the initial stage, development stage, outbreak stage and decline stage of Ulva prolifera are collected by using satellites.

Images