Method for identifying invasion areas in mangrove ecological regions based on hyperspectral images

Abstract

The present invention belongs to the technical field of intelligent monitoring and remote sensing mapping, and proposes a method for identifying invasion areas in mangrove ecological areas based on hyperspectral images, collects spectral cubes of mangrove ecological areas, grids the mangrove ecological areas to form grid units, maps the grid units to spectral cubes, reads spectral angle binaries of grid units from spectral cubes, performs time series differential analysis on spectral angle binaries to form invasion analysis values, and obtains invasion measurement results according to the invasion analysis values. The method uses hyperspectral imaging technology to accurately monitor the growth density and distribution of niche-occupying plants, identifies microscopic changes in plant growth and achieves early warning, predicts and identifies the occurrence of plant diffusion to reduce the risk of local ecological damage to mangroves, provides data support for the long-term management and protection of mangrove areas, and promotes the health and stability of the ecosystem.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent monitoring and remote sensing mapping, and in particular relates to a method for identifying an invasion area of ​​a mangrove ecological region based on a hyperspectral image. Background Art

[0002] Mangroves are an important part of coastal wetland ecosystems, not only providing habitats for a variety of organisms, but also playing key ecological functions in terms of windbreak, carbon sequestration and water purification. However, the spread of invasive plants such as Spartina alterniflora, Mikania micrantha, Sonneratia apetala and Derris ternata is encroaching on the growth space of mangroves and interfering with the species diversity and community structure of benthic organisms by changing the physical and chemical properties of the soil, thus leading to a reduction in the area of ​​mangroves and a decline in their ecological service functions.

[0003] At present, the monitoring of specific plants in mangroves mainly relies on ground surveys and remote sensing imaging technology. Although ground surveys can provide intuitive data, due to their low efficiency, high cost and slow data update, they are difficult to meet the precise monitoring needs of niche-occupying plants that are rapidly spreading in mangrove ecological areas. Traditional remote sensing imaging technology mainly collects the morphology and limited spectral information of ground targets. However, due to the limitations of image resolution and intelligent recognition algorithms, small and medium-sized niche-occupying plants are often ignored, which makes it impossible to detect plant invasions in time during regular monitoring, which is not conducive to early prevention and control.

[0004] With the emergence of hyperspectral imaging technology, its high spectral resolution can finely detect the spectral characteristics of objects. This technology has significantly improved the monitoring capabilities of specific plants in mangrove ecological areas. Through real-time or near real-time data collection and processing, it can quickly respond to the spread of niche-occupying plants, especially in early detection and accurate identification, providing more effective technical support for the scientific management and protection of mangroves.

[0005] Therefore, a mangrove ecological area invasion area identification method based on hyperspectral images is needed to accurately identify and distinguish the spectral characteristics of niche-occupying plants, combined with intelligent algorithms to adapt to rapidly changing monitoring needs and reduce misjudgments due to excessive sensitivity. Summary of the invention

[0006] The purpose of the present invention is to propose a method for identifying invasive areas in mangrove ecological regions based on hyperspectral images to solve one or more technical problems existing in the prior art and at least provide a beneficial option or create conditions.

[0007] In order to achieve the above object, according to one aspect of the present invention, a method for identifying an invasion area of ​​a mangrove ecological region based on a hyperspectral image is provided, the method comprising the following steps:

[0008] S100, collecting spectral cubes of mangrove ecoregions;

[0009] S200, gridding the mangrove ecoregion to form grid cells and mapping the grid cells to the spectral cube;

[0010] S300, read the spectral angle binary of the grid cell from the spectral cube;

[0011] S400, performing time-series difference analysis on the spectral angle binary group to form an intrusion analysis value;

[0012] S500, obtaining an intrusion measurement result according to the intrusion analysis value;

[0013] S600, identifies the types of invasive species in mangrove ecoregions by combining the invasion determination results of each invasive species.

[0014] Furthermore, in step S100, the method for collecting the spectral cube of the mangrove ecological zone is: collecting the hyperspectral image of the mangrove ecological zone by a satellite or an unmanned aerial vehicle equipped with a hyperspectral imager; preprocessing the hyperspectral image obtained after scanning by the hyperspectral imager, the preprocessing step includes using the FLAASH atmospheric correction model to compensate for the atmospheric effect, and performing geometric correction processing through the DEM digital elevation model, and forming a spectral cube by the hyperspectral images corresponding to each band.

[0015] The atmospheric correction process is used to correct the data distortion caused by atmospheric absorption and scattering effects, and the geometric correction is used to correct the geometric deformation of the image during the projection process to ensure the spatial geometric relationship of the image and the matching accuracy of the GPS position;

[0016] Preprocessing also includes using GIS software to crop the hyperspectral image so that the data collection area of ​​the hyperspectral image is constrained to a limited or preset mangrove ecological zone.

[0017] The hyperspectral imager is a Nano-Hyperspec full-band hyperspectral imager with a wavelength range of 400-1000nm, 270 spectral channels, 640 spatial channels, a spectral sampling interval of 2.2nm / pixel, and a maximum frame rate of 350fps.

[0018] Further, in step S300, the method of reading the spectral angle binary of the grid unit from the spectral cube is: spatially aggregating the spectral information of the same grid unit in the spectral cube to form a spectral feature vector of the grid unit; performing spectral angle SAM analysis on the spectral feature vector of the grid unit to obtain a spectral angle binary; wherein the spectral angle binary includes a balanced difference and a representative difference; wherein the balanced difference is the difference between the balanced spectral angle of mangroves and the invasion spectral angle, and the representative difference is the difference between the representative spectral angle of mangroves and the invasion spectral angle; the balanced spectral angle of mangroves is the weighted average of the SAM analysis results of each mangrove species, and the weight of any mangrove species is obtained by pre-setting in advance, or is set to the frequency value of the mangrove species appearing in all grid cells; the representative spectral angle of mangroves is the maximum value among the SAM analysis results of each mangrove species; and the invasion spectral angle is the SAM analysis result of the invasive species.

[0019] If the weight of any mangrove species is not obtained by presetting in advance, it is set to the frequency value of the mangrove species in all grid cells;

[0020] One natural day is taken as a measuring point, and a spectral angle binary pair is collected at each measuring point.

[0021] Spatial aggregation refers to calculating the average value of each pixel value in the same band within the grid unit as the representative value of the grid in the band. It is a dimensionality reduction technology that simplifies the complexity of the data.

[0022] The invasive species is any one of Spartina alterniflora, Mikania micrantha, Sonneratia apetala, Derris dasyphylla or Alternanthera philoxeroides.

[0023] Mangrove species include two or more of Aegiceras corniculata, Avicennia marina, Rhus chinensis, Bruguiera gymnorrhiza, Pongamia chinensis, Kandelia candel and Rhizophora rubra.

[0024] The species spectrum benchmark for spectral angle analysis uses the typical spectrum database in the ECOSTRESS Spectral Library platform.

[0025] Furthermore, in step S400, the method of performing time-series difference analysis on the spectral angle binary group to form an intrusion analysis value is: defining the maximum value of the balanced difference of all grid cells at any measuring point as the difference peak value, and the difference peak value corresponds to the grid cell at which the peak value critical occurs at the measuring point;

[0026] Iterate in reverse time direction from the current moment, and the iteration variables include the peak critical benchmark, and mark the grid unit where the peak critical occurs under the current measurement point as critical; the initial value of the peak critical benchmark is the maximum value among the current equilibrium difference degrees;

[0027] In any iteration, the peak value of the difference under the iterative measurement point is compared with the peak critical benchmark, the peak critical benchmark is updated to the minimum value of the two, and each grid unit whose peak value of the difference under the iterative measurement point is greater than the peak critical benchmark is marked as critical;

[0028] The iteration end condition is that all grid cells have at least one critical mark, and the measurement point that meets the iteration end condition is the last measurement point; each measurement point between the current measurement point and the last measurement point constitutes the analytical interval;

[0029] In the analytical interval, for any grid unit, the first difference component Dst of any measuring point is the difference between the equilibrium difference degree of the measuring point and the difference degree peak value corresponding to the first critical mark point of the reverse time search, and the second difference component is the difference between the representative difference degree of the measuring point and the difference degree peak value corresponding to the first critical mark point of the reverse time search. Then the difference component PD is the element with the maximum absolute value between the first difference component and the second difference component;

[0030] If the difference value of a grid unit at a measuring point is taken from the first difference value, the measuring point is defined as the strong difference position of this grid unit. The intrusion analysis value Trso of the grid unit is calculated according to each strong difference position. The calculation method is:

[0031] ;

[0032] Where rvt is the difference reversal rate. When the product of the strong difference bit and the difference component of the previous measurement point is less than 0, the strong difference bit is defined to be flipped. The ratio of the number of flipped strong difference bits to the total number of strong difference bits is the difference reversal rate. Dst.err is the standard deviation of all first difference components in the analytical interval, KR represents the total number of strong difference bits, i1 is the accumulated variable, and PD i1 represents the difference of the i1th strong difference bit, rt.bia(Dst i1 ) represents the ratio of the first difference component of the i1th strong difference position to the average value of the first difference components corresponding to each grid unit under the measuring point.

[0033] Since the invasion analysis value depends on the strong difference position obtained by calculating and comparing the first difference component and the second difference component in the analysis interval, it can effectively screen out the measurement points with significant characteristics in each grid unit, so as to quantify the diffusion phenomenon of niche occupying plants and be used for early warning of plant invasion in the mangrove ecological area. However, in the process of obtaining the difference component, the element with the maximum absolute value is selected, and the measurement of other lower difference values ​​is ignored, so that the sensitivity of the data will form a landslide when the time distribution density of the strong difference position is greatly different, and then the problem of decreased sensitivity to the invasion analysis value will occur. However, the existing technology cannot effectively compensate for this phenomenon of decreased sensitivity. In order to eliminate this influence, the present invention proposes a more preferred solution as follows:

[0034] Furthermore, in step S400, the invasion analysis value is input into the elastic network regression model to identify the species invasion area. The method is as follows: define the neighboring area of ​​any grid cell as a 3*3 or 5*5 neighborhood block intercepted with the grid cell as the center; obtain the minimum value of the representative difference in any grid cell and record it as the representative difference base value, and then the ratio of the representative difference to the representative difference base value of the corresponding grid cell is recorded as the representative overflow rate SRI_Rt;

[0035] Take any grid unit as the current grid unit. In the current grid unit, if the representative excess rate of the current measuring point is less than the minimum value of the first representative excess rate in the counterclockwise direction, the grid unit of the measuring point is recorded as the first characteristic unit.

[0036] The maximum value of the representative overflow rate of each first characteristic unit in the current measuring point is calculated, and the root mean square value of the calculated maximum value and the balanced difference degree of each grid unit is recorded as the balanced overflow value Fsim;

[0037] The balanced overflow value used in the overflow rate is calculated by combining this maximum anomaly value with the balanced difference of all measuring points in each grid unit. It represents the overall deviation between the balanced characteristics of all grid units and the most abnormal unit, which is equivalent to enhancing the contrast between the balance of spectral characteristics and local anomalies in a grid unit.

[0038] Define the average value of the balanced overflow phase values ​​obtained by all grid cells at the current measuring point as Eqo; the balanced overflow phase values ​​obtained by all grid cells at the current measuring point constitute a sequence denoted as ls.Fge, and set the feature selection condition as: Fsimi≥min(ls.Fge)·SRI_Rti, where min() is the minimum function, Fsimi and SRI_Rti represent the balanced overflow phase value and the overflow rate of the i-th grid cell at the current measuring point respectively; when a grid cell meets the feature selection condition and does not belong to the first feature cell, the grid cell is defined as the second feature cell at the current measuring point;

[0039] The ratio of the equilibrium overflow value of each second characteristic unit at the same measuring point to the number of second characteristic units at the measuring point is recorded as the overflow opportunity Rud;

[0040] At the current measuring point, define the median value of the overflow opportunity in the adjacent area of ​​any second feature unit as the regional overflow opportunity, search for other second feature units that are closest to the current second feature unit and whose overflow opportunity is greater than the regional overflow opportunity of the current second feature unit, and record the distance between the two as the overflow opportunity radius; take any second feature unit as the center and the overflow opportunity radius as the radius to form a circle as the overflow opportunity area, calculate the average distance from each first feature unit in the overflow opportunity area to the current second feature unit, and record it as the inter-element analytical coefficient Inc;

[0041] The obtained inter-element analysis coefficient is a way to measure the spatial expansion range of the overflow phenomenon. A smaller average distance means that the overflow phenomenon is more limited, that is, the second characteristic unit has a smaller range of influence when performing differential analysis at the current measuring point. At the same time, considering the relationship between the overflow phenomenon and the regional anomaly characteristics in the subsequent calculation process, this inter-element analysis coefficient can provide information about the response intensity of the local anomaly in space that affects the surrounding area, that is, a smaller average distance may imply a strong impact of the local anomaly.

[0042] The intrusion analysis value Trso is calculated based on the balanced overflow phase value. If the grid unit is the first characteristic unit, the intrusion analysis value is not calculated; if the grid unit does not belong to the second characteristic unit at the current measurement point, the overflow opportunity area when it was the second characteristic unit most recently is used; the most recent time as the second characteristic unit refers to the first measurement point in the reverse time direction where the grid unit belongs to the second characteristic unit, and the intrusion analysis value is Trso:

[0043] ;

[0044] Where j1 is the accumulated variable, Trso j1 is the intrusion analysis value of the current grid cell at the j1th measurement point, HSA(j1) is the average value of the equilibrium overflow value of the first characteristic unit in the overflow opportunity area corresponding to the grid cell at the j1th measurement point, SRI_Rt j1 is the representative overflow rate of the current grid cell at the j1th measurement point, MSRI_Rt j1 is the median value of the representative overflow rate from the grid unit under the j1th measuring point to the first second characteristic unit in the reverse time direction, Inc j1 is the inter-element analytical coefficient of the overflow opportunity area corresponding to the grid unit under the j1th measuring point, and lg() is a logarithmic function with base 10.

[0045] In the above method, when the maximum and minimum values ​​are screened, the default time constraint is the period from the current natural day to the 365 natural days before it;

[0046] Furthermore, in step S500, the method for obtaining the intrusion measurement result according to the intrusion analysis value is: calculate the time series feature quantity from all the intrusion analysis values ​​obtained from the history of any grid unit through the tsfresh library, and construct each time series feature quantity into a sequence recorded as a grid unit feature sequence; record the number of grid units and the number of time series feature quantities as m and n respectively, and construct a feature matrix STMX with a dimension of m×n according to the grid unit feature sequence of each grid unit; use the moving average coefficient as the target feature input to the pre-trained elastic network regression model to obtain the prediction result of the target feature corresponding to each grid unit, and when the prediction result is true, the intrusion measurement result of the grid unit is true, otherwise the intrusion measurement result is false.

[0047] In this process, the variation of the grid unit is located by the elastic network regression algorithm, and based on the temporal variation characteristics of each grid unit itself as a feature, the prediction results of the model are used to determine whether the node has mutated. The threshold of the moving average coefficient is predefined in the pre-trained elastic network regression model, and the threshold is set between 0.10 and 0.20. When the moving average coefficient exceeds the threshold, the return result is true, otherwise it is false; when the return result is true, the temporal characteristics of the grid unit indicate that the grid unit has species invasion. The occurrence of species invasion refers to the obvious growth trend of the invasive species that needs to be identified in the current grid unit in the area.

[0048] Furthermore, it is characterized in that the time series feature quantity includes one or more of an autoregressive coefficient, a moving average coefficient, an ADF test statistic, a sample entropy, an FFT component, a linear trend slope, an autocorrelation coefficient and a partial autocorrelation coefficient.

[0049] Furthermore, in step S600, the method for identifying the types of invasive species in the mangrove ecological zone in combination with the invasion determination results of each invasive species is: obtaining the invasive species for which each invasion determination result is true as the determined invasive species, and the types of invasive species in the mangrove ecological zone in any grid unit include all the determined invasive species obtained in the grid unit.

[0050] Preferably, all undefined variables in the present invention, if not clearly defined, can be manually set thresholds.

[0051] The present invention also provides a mangrove ecological region invasive species identification system based on hyperspectral images, the mangrove ecological region invasive species identification system based on hyperspectral images comprising: a processor, a memory and a computer program stored in the memory and executable on the processor, the processor implementing the steps in the mangrove ecological region invasive region identification method based on hyperspectral images when executing the computer program, the mangrove ecological region invasive species identification system based on hyperspectral images can be run in computing devices such as desktop computers, laptop computers, PDAs and cloud data centers, and the executable system may include, but is not limited to, a processor, a memory, and a server cluster, and the processor executes the computer program to run in the following system units:

[0052] A data collection unit to collect spectral cubes of the mangrove ecoregion;

[0053] Grid mapping unit, which forms grid cells by gridding the mangrove eco-region and maps the grid cells to the spectral cube;

[0054] A data reading unit, used for reading the spectral angle doublet of the grid cell from the spectral cube;

[0055] A time series difference analysis unit, used for performing time series difference analysis on the spectral angle binary group to form an intrusion analysis value;

[0056] An intrusion determination unit, used for obtaining an intrusion determination result according to the intrusion analysis value;

[0057] The invasive species identification unit is used to identify the types of invasive species in the mangrove ecological zone by combining the invasion determination results of each invasive species.

[0058] The beneficial effects of the present invention are as follows: the method for monitoring changes in mangrove ecological regions of the present invention utilizes hyperspectral imaging technology to accurately collect remote sensing image data of niche-occupying plants, and realizes accurate monitoring of the growth density and distribution range of niche-occupying plants by analyzing differences in plant spectral characteristics, quantifies the diffusion and invasion phenomena of niche-occupying plants, and provides early warning for plant invasions in mangrove ecological regions, especially with superior response timeliness, thereby controlling the diffusion of niche-occupying plants and reducing their damage to mangrove ecology. The monitoring method of the present invention can also provide a scientific mathematical basis for the long-term management and protection of mangrove ecological regions, and serve as a reference feature for future species invasion type data analysis, ultimately promoting the overall health and stability of the mangrove ecosystem. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] The above and other features of the present invention will become more obvious by describing in detail the embodiments shown in the accompanying drawings. The same reference numerals in the accompanying drawings of the present invention represent the same or similar elements. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other accompanying drawings can be obtained based on these accompanying drawings without creative work. In the accompanying drawings:

[0060] Figure 1 Shown is a flow chart of the method for identifying invasion areas in mangrove ecoregions based on hyperspectral images;

[0061] Figure 2 Shown is the structure diagram of the invasive species identification system in mangrove ecological areas based on hyperspectral images;

[0062] Figure 3 The figure shows the area where Alternaria alternifolia invasion occurred, which was identified by the hybrid decision tree classification model in the comparative example;

[0063] Figure 4 Shown is the area where Alternaria alternifolia invasion occurred, as identified in Example 1;

[0064] Figure 5 Shown are the areas where Alternaria alternifolia invasion has occurred using the identification obtained in Example 2. DETAILED DESCRIPTION

[0065] The following will be combined with the embodiments and drawings to clearly and completely describe the concept, specific structure and technical effects of the present invention, so as to fully understand the purpose, scheme and effect of the present invention. It should be noted that the embodiments and features in the embodiments of this application can be combined with each other without conflict.

[0066] like Figure 1 The figure shows the flow chart of the method for identifying the invasion area of ​​mangrove ecological regions based on hyperspectral images. Figure 1 The following describes a method for identifying an invasion area of ​​a mangrove ecological region based on a hyperspectral image according to an embodiment of the present invention. The method comprises the following steps:

[0067] Example 1

[0068] This embodiment uses mangroves in a coastal wetland scenic area in a certain area as the identification object.

[0069] S100, collecting spectral cubes of mangrove ecoregions;

[0070] S200, gridding the mangrove ecoregion to form grid cells and mapping the grid cells to the spectral cube;

[0071] S300, read the spectral angle binary of the grid cell from the spectral cube;

[0072] S400, performing time-series difference analysis on the spectral angle binary group to form an intrusion analysis value;

[0073] S500, obtaining an intrusion measurement result according to the intrusion analysis value;

[0074] S600, identifies the types of invasive species in mangrove ecoregions by combining the invasion determination results of each invasive species.

[0075] Furthermore, in step S100, the method for collecting the spectral cube of the mangrove ecological zone is: collecting the hyperspectral image of the mangrove ecological zone by a satellite or an unmanned aerial vehicle equipped with a hyperspectral imager; preprocessing the hyperspectral image obtained after scanning by the hyperspectral imager, the preprocessing step includes using the FLAASH atmospheric correction model to compensate for the atmospheric effect, and performing geometric correction processing through the DEM digital elevation model, and forming a spectral cube by the hyperspectral images corresponding to each band.

[0076] The atmospheric correction process is used to correct the data distortion caused by atmospheric absorption and scattering effects, and the geometric correction is used to correct the geometric deformation of the image during the projection process to ensure the spatial geometric relationship of the image and the matching accuracy of the GPS position;

[0077] Preprocessing also includes using GIS software to crop the hyperspectral image so that the data collection area of ​​the hyperspectral image is constrained to a limited or preset mangrove ecological zone.

[0078] Furthermore, in step S200, the mangrove ecological area is gridded to form grid units, and the method of mapping the grid units uses a vector creation tool of ArcGIS or QGIS software to create a grid.

[0079] Further, in step S300, the method of reading the spectral angle binary of the grid unit from the spectral cube is: spatially aggregating the spectral information of the same grid unit in the spectral cube to form a spectral feature vector of the grid unit; performing spectral angle SAM analysis on the spectral feature vector of the grid unit to obtain a spectral angle binary; wherein the spectral angle binary includes a balanced difference and a representative difference; wherein the balanced difference is the difference between the balanced spectral angle of mangroves and the invasion spectral angle, and the representative difference is the difference between the representative spectral angle of mangroves and the invasion spectral angle; the balanced spectral angle of mangroves is the weighted average of the SAM analysis results of each mangrove species, and the weight of any mangrove species is obtained by pre-setting in advance, or is set to the frequency value of the mangrove species appearing in all grid cells; the representative spectral angle of mangroves is the maximum value among the SAM analysis results of each mangrove species; and the invasion spectral angle is the SAM analysis result of the invasive species.

[0080] The invasive species is set as Spartina alterniflora, and the mangrove species include Avicennia marina, Bruguiera gymnorrhiza, Pongamia chinensis, Kandelia candel and Rhizophora rubrum.

[0081] Furthermore, in step S400, the method of performing time-series difference analysis on the spectral angle binary group to form an intrusion analysis value is: defining the maximum value of the balanced difference of all grid cells at any measuring point as the difference peak value, and the difference peak value corresponds to the grid cell at which the peak value critical occurs at the measuring point;

[0082] Iterate in reverse time direction from the current moment, and the iteration variables include the peak critical benchmark, and mark the grid unit where the peak critical occurs under the current measurement point as critical; the initial value of the peak critical benchmark is the maximum value among the current equilibrium difference degrees;

[0083] In any iteration, the peak value of the difference under the iterative measurement point is compared with the peak critical benchmark, the peak critical benchmark is updated to the minimum value of the two, and each grid unit whose peak value of the difference under the iterative measurement point is greater than the peak critical benchmark is marked as critical;

[0084] The iteration end condition is that all grid cells have at least one critical mark, and the measurement point that meets the iteration end condition is the last measurement point; each measurement point between the current measurement point and the last measurement point constitutes the analytical interval;

[0085] In the analytical interval, for any grid unit, the first difference component Dst of any measuring point is the difference between the equilibrium difference degree of the measuring point and the difference degree peak value corresponding to the first critical mark point of the reverse time search, and the second difference component is the difference between the representative difference degree of the measuring point and the difference degree peak value corresponding to the first critical mark point of the reverse time search. Then the difference component PD is the element with the maximum absolute value between the first difference component and the second difference component;

[0086] If the difference value of a grid unit at a measuring point is taken from the first difference value, the measuring point is defined as the strong difference position of this grid unit. The intrusion analysis value Trso of the grid unit is calculated according to each strong difference position. The calculation method is:

[0087] ;

[0088] Where rvt is the difference reversal rate. When the product of the strong difference bit and the difference component of the previous measurement point is less than 0, the strong difference bit is defined to be flipped. The ratio of the number of flipped strong difference bits to the total number of strong difference bits is the difference reversal rate. Dst.err is the standard deviation of all first difference components in the analytical interval, KR represents the total number of strong difference bits, i1 is the accumulated variable, and PD i1 represents the difference of the i1th strong difference bit, rt.bia(Dst i1 ) represents the ratio of the first difference component of the i1th strong difference position to the average value of the first difference components corresponding to each grid unit under the measuring point.

[0089] Furthermore, in step S500, the method for obtaining the intrusion measurement result according to the intrusion analysis value is: calculate the time series feature quantity from all the intrusion analysis values ​​obtained from the history of any grid unit through the tsfresh library, and construct each time series feature quantity into a sequence recorded as a grid unit feature sequence; record the number of grid units and the number of time series feature quantities as m and n respectively, and construct a feature matrix STMX with a dimension of m×n according to the grid unit feature sequence of each grid unit; use the moving average coefficient as the target feature input to the pre-trained elastic network regression model to obtain the prediction result of the target feature corresponding to each grid unit, and when the prediction result is true, the intrusion measurement result of the grid unit is true, otherwise the intrusion measurement result is false.

[0090] Furthermore, it is characterized in that the time series feature quantities include autoregressive coefficient, moving average coefficient, ADF test statistic, sample entropy, FFT component, linear trend slope, autocorrelation coefficient and partial autocorrelation coefficient.

[0091] Furthermore, in step S600, the method for identifying the types of invasive species in the mangrove ecological zone in combination with the invasion determination results of each invasive species is: for any grid unit in the mangrove ecological zone, if the invasion determination result of Spartina alterniflora is true, then Spartina alterniflora is identified as the invasive species in the grid unit as the determined invasive species, and each grid unit where Spartina alterniflora invasion occurs is filled with black on the image to obtain each area where Spartina alterniflora invasion occurs.

[0092] Figure 4 The black grid area shown is the area where the invasion of Alternaria alternifolia occurs, as identified in Example 1;

[0093] Example 2

[0094] Embodiment 2 uses the same method as Embodiment 1 to identify the intrusion area, except that, in step S400, the method of performing temporal difference analysis on the spectral angle binary to form the intrusion analysis value is replaced by: defining the adjacent area of ​​any grid unit as a 3*3 or 5*5 neighborhood block intercepted with the grid unit as the center; obtaining the minimum value of the representative difference in any grid unit and recording it as the representative difference base value, and then the ratio of the representative difference to the representative difference base value of the corresponding grid unit is recorded as the representative overflow rate SRI_Rt;

[0095] Take any grid unit as the current grid unit. In the current grid unit, if the representative excess rate of the current measuring point is less than the minimum value of the first representative excess rate in the counterclockwise direction, the grid unit of the measuring point is recorded as the first characteristic unit.

[0096] The maximum value of the representative overflow rate of each first characteristic unit in the current measuring point is calculated, and the root mean square value of the calculated maximum value and the balanced difference degree of each grid unit is recorded as the balanced overflow value Fsim;

[0097] Define the average value of the balanced overflow values ​​obtained by all grid cells at the current measuring point as Eqo; the sequence of balanced overflow values ​​obtained by all grid cells at the current measuring point is recorded as ls.Fge, and the feature selection condition is set as: Fsim i ≥min(ls.Fge)·SRI_Rt i , where min() is the minimum function, Fsim i and SRI_Rt i Respectively represent the equilibrium overflow value and the overflow rate of the i-th grid unit at the current measuring point; when a grid unit meets the feature selection conditions and does not belong to the first feature unit, the grid unit is defined as the second feature unit at the current measuring point;

[0098] The ratio of the equilibrium overflow value of each second characteristic unit at the same measuring point to the number of second characteristic units at the measuring point is recorded as the overflow opportunity Rud;

[0099] At the current measuring point, define the median value of the overflow opportunity in the adjacent area of ​​any second feature unit as the regional overflow opportunity, search for other second feature units that are closest to the current second feature unit and whose overflow opportunity is greater than the regional overflow opportunity of the current second feature unit, and record the distance between the two as the overflow opportunity radius; take any second feature unit as the center and the overflow opportunity radius as the radius to form a circle as the overflow opportunity area, calculate the average distance from each first feature unit in the overflow opportunity area to the current second feature unit, and record it as the inter-element analytical coefficient Inc;

[0100] The intrusion analysis value Trso is calculated based on the balanced overflow phase value. If the grid unit is the first characteristic unit, the intrusion analysis value is not calculated; if the grid unit does not belong to the second characteristic unit at the current measurement point, the overflow opportunity area when it was the second characteristic unit most recently is used; the most recent time as the second characteristic unit refers to the first measurement point in the reverse time direction where the grid unit belongs to the second characteristic unit, and the intrusion analysis value is Trso:

[0101] ;

[0102] Where j1 is the accumulated variable, Trso j1 is the intrusion analysis value of the current grid cell at the j1th measurement point, HSA(j1) is the average value of the equilibrium overflow value of the first characteristic unit in the overflow opportunity area corresponding to the grid cell at the j1th measurement point, SRI_Rt j1is the representative overflow rate of the current grid cell at the j1th measurement point, MSRI_Rt j1 is the median value of the representative overflow rate from the grid unit under the j1th measuring point to the first second characteristic unit in the reverse time direction, Inc j1 is the inter-element analytical coefficient of the overflow opportunity area corresponding to the grid unit under the j1th measuring point, and lg() is a logarithmic function with base 10.

[0103] like Figure 5 The black grid area shown is the area where the invasion of Alternaria alternifolia occurs using the identification obtained in Example 2.

[0104] Comparative Example

[0105] The spectral cube collected at the current moment is input into the pre-trained hybrid decision tree classification model, wherein the maximum depth of the hybrid decision tree is 15, the minimum sample split number is 8, the splitting standard is gini type, and the minimum number of leaf node samples is 5; the hybrid decision tree classification model is used to obtain the invasion probability value of the Alternaria japonica corresponding to each pixel in the mangrove ecological area, and the mangrove ecological area is mapped using the same grid unit as in Example 1. When the invasion probability value of the Alternaria japonica in the grid unit is higher than 80%, it is defined that the invasive species in the grid unit include the Alternaria japonica.

[0106] like Figure 3 As shown, the black grid area is the area where the invasion of Alternaria alternifolia occurs, which is identified by the hybrid decision tree classification model in the comparative example;

[0107] According to the recognition results of the comparative examples and the embodiments, it can be found that the traditional static recognition method is very prone to over-recognition problems caused by overfitting, thereby misleadingly identifying areas where no real invasion has occurred, which brings great difficulty to the further prevention and control tasks. This is because the prevention and control of invasive species in a specific area often requires periodic repetition of invasive plant discovery and regular review and other inspections. Compared with the comparative examples, the embodiments more accurately identify the invasion location, and greatly reduce the risk of misleading data when the source data is collected at a single time through the time series analysis method. The principle is actually to apply the performance of the diffusion invasion of niche occupying plants in terms of persistence, so the accuracy of recognition can be ensured.

[0108] The embodiment of the present invention provides a system for identifying invasive species in mangrove ecological areas based on hyperspectral images, such as Figure 2What is shown is a structural diagram of the invasive species identification system for mangrove ecological areas based on hyperspectral images of the present invention. The invasive species identification system for mangrove ecological areas based on hyperspectral images of this embodiment includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the above-mentioned embodiment of the method for identifying invasive areas of mangrove ecological areas based on hyperspectral images are implemented.

[0109] The system comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to run in the following units of the system:

[0110] A data collection unit to collect spectral cubes of the mangrove ecoregion;

[0111] Grid mapping unit, which forms grid cells by gridding the mangrove eco-region and maps the grid cells to the spectral cube;

[0112] A data reading unit, used for reading the spectral angle doublet of the grid cell from the spectral cube;

[0113] A time series difference analysis unit, used for performing time series difference analysis on the spectral angle binary group to form an intrusion analysis value;

[0114] An intrusion determination unit, used for obtaining an intrusion determination result according to the intrusion analysis value;

[0115] The invasive species identification unit is used to identify the types of invasive species in the mangrove ecological zone by combining the invasion determination results of each invasive species.

[0116] The mangrove ecological region invasive species identification system based on hyperspectral images can be run on computing devices such as desktop computers, laptops, PDAs, and cloud servers. The mangrove ecological region invasive species identification system based on hyperspectral images, the executable system may include, but is not limited to, processors and memories. Those skilled in the art will appreciate that the examples are merely examples of mangrove ecological region invasive species identification systems based on hyperspectral images, and do not constitute a limitation on mangrove ecological region invasive species identification systems based on hyperspectral images, and may include more or fewer components than the examples, or a combination of certain components, or different components, for example, the mangrove ecological region invasive species identification system based on hyperspectral images may also include input and output devices, network access devices, buses, and the like.

[0117] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the operation system of the mangrove ecological region invasive species identification system based on hyperspectral images, and uses various interfaces and lines to connect various parts of the entire mangrove ecological region invasive species identification system based on hyperspectral images.

[0118] The memory can be used to store the computer program and / or module, and the processor realizes various functions of the mangrove ecological area invasive species identification system based on hyperspectral images by running or executing the computer program and / or module stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data created according to the use of the mobile phone (such as audio data, a phone book, etc.), etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, 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), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0119] Although the description of the present invention has been quite detailed and has been described in particular with respect to several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, so as to effectively cover the intended scope of the present invention. In addition, the present invention is described above with the embodiments foreseeable by the inventors, and its purpose is to provide a useful description, and those non-substantial changes to the present invention that are not currently foreseen may still represent equivalent changes of the present invention.

Claims

1. A method for identifying invasion areas in mangrove ecological regions based on hyperspectral images, characterized in that: The method comprises the following steps: S100, collecting spectral cubes of mangrove ecoregions; S200, gridding the mangrove ecoregion to form grid cells and mapping the grid cells to the spectral cube; S300, read the spectral angle binary of the grid cell from the spectral cube; S400, performing time-series difference analysis on the spectral angle binary group to form an intrusion analysis value; S500, obtaining an intrusion measurement result according to the intrusion analysis value; S600, combining the invasion determination results of each invasive species to identify the types of invasive species in the mangrove ecoregion; Wherein, in step S300, the method for reading the spectral angle binary of the grid unit from the spectral cube is: spatially aggregating the spectral information of the same grid unit in the spectral cube to form a spectral feature vector of the grid unit; performing spectral angle SAM analysis on the spectral feature vector of the grid unit to obtain a spectral angle binary; wherein the spectral angle binary includes a balanced difference and a representative difference; wherein the balanced difference is the difference between the balanced spectral angle of mangrove and the invasion spectral angle, and the representative difference is the difference between the representative spectral angle of mangrove and the invasion spectral angle; the balanced spectral angle of mangrove is a weighted average of the SAM analysis results of each mangrove species, and the weight of any mangrove species is obtained by presetting in advance, or is set to the frequency value of the mangrove species appearing in all grid cells; the representative spectral angle of mangrove is the maximum value among the SAM analysis results of each mangrove species; the invasion spectral angle is the SAM analysis result of the invasive species, and the invasive species is one of Spartina alterniflora, Mikania micrantha, Sonneratia apetala, Derris dasyphylla, or Alternanthera philoxeroides; In step S400, the method of performing time-series differential analysis on the spectral angle binary group to form an intrusion analysis value is as follows: define the maximum value of the balanced difference of all grid cells at any measuring point as the difference peak value, and the grid cell corresponding to the difference peak value has a peak criticality at the measuring point; iterate in the reverse time direction from the current moment, and the iteration variable includes the peak criticality benchmark, and the grid cell having a peak criticality at the current measuring point is marked as a criticality; the initial value of the peak criticality benchmark is the maximum value of each current balanced difference; In any iteration, the difference peak value under the iterative measurement point is compared with the peak critical benchmark, and the peak critical benchmark is updated to the minimum value of the two. Each grid unit whose difference peak value under the iterative measurement point is greater than the peak critical benchmark is marked as critical; the iteration end condition is that all grid units have at least one critical mark, and the measurement point that meets the iteration end condition is the last measurement point; each measurement point between the current measurement point and the last measurement point constitutes the analytical interval; In the analysis interval, for any grid unit, the first difference component Dst of any measuring point is the difference between the equilibrium difference degree of the measuring point and the difference peak value corresponding to the first critical mark point of the reverse time search, and the second difference component is the difference between the representative difference degree of the measuring point and the difference peak value corresponding to the first critical mark point of the reverse time search. The difference component PD is the element with the maximum absolute value between the first difference component and the second difference component. If the difference component of the grid unit at a measuring point is taken from the first difference component, the measuring point is defined as the strong difference position of this grid unit, and the intrusion analysis value Trso of the grid unit is calculated according to each strong difference position. In step S500, the method for obtaining the intrusion measurement result according to the intrusion analysis value is: calculate the time series feature quantity from all the intrusion analysis values ​​obtained from the history of any grid unit through the tsfresh library, and construct each time series feature quantity into a sequence recorded as a grid unit feature sequence; record the number of grid units and the number of time series feature quantities as m and n respectively, and construct a feature matrix STMX with a dimension of m×n according to the grid unit feature sequence of each grid unit; use the moving average coefficient as the target feature input to the pre-trained elastic network regression model to obtain the prediction result of the target feature corresponding to each grid unit, and when the prediction result is true, the intrusion measurement result of the grid unit is true, otherwise the intrusion measurement result is false.

2. The method for identifying mangrove ecological region invasion areas based on hyperspectral images according to claim 1 is characterized in that: In step S100, the method for collecting the spectral cube of the mangrove ecological zone is: collecting a hyperspectral image of the mangrove ecological zone by using a satellite or a drone equipped with a hyperspectral imager; The hyperspectral images obtained after scanning by the hyperspectral imager are preprocessed. The preprocessing steps include using the FLAASH atmospheric correction model to compensate for atmospheric effects, performing geometric correction through the DEM digital elevation model, and constructing a spectral cube from the hyperspectral images corresponding to each band.

3. The method for identifying mangrove ecological region invasion areas based on hyperspectral images according to claim 1 is characterized in that: In step S400, the method of performing time-series difference analysis on the spectral angle binary group to form the intrusion analysis value can be replaced by: defining the adjacent area of ​​any grid cell as a 3*3 or 5*5 neighborhood block intercepted with the grid cell as the center; obtaining the minimum value of the representative difference in any grid cell and recording it as the representative difference base value, and then recording the ratio of the representative difference to the representative difference base value of its corresponding grid cell as the representative overflow rate SRI_Rt; taking any grid cell as the current grid cell, in the current grid cell, if the representative overflow rate of the current measuring point is less than the minimum value of the first representative overflow rate in the counterclockwise direction, then recording the grid cell of the measuring point as the first characteristic unit; The maximum value of the representative overflow rate of each first characteristic unit in the current measuring point is calculated, and the root mean square value of the calculated maximum value and the balanced difference degree of each grid unit is recorded as the balanced overflow value Fsim; Define the average value of the balanced overflow values ​​obtained by all grid cells at the current measuring point as Eqo; the sequence of balanced overflow values ​​obtained by all grid cells at the current measuring point is recorded as ls.Fge, and the feature selection condition is set as: Fsim i ≥min(ls.Fge)·SRI_Rt i , where min() is the minimum function, Fsim i and SRI_Rt i Respectively represent the equilibrium overflow value and the overflow rate of the i-th grid unit at the current measuring point; when a grid unit meets the feature selection conditions and does not belong to the first feature unit, the grid unit is defined as the second feature unit at the current measuring point; The ratio of the equilibrium overflow value of each second characteristic unit at the same measuring point to the number of second characteristic units at the measuring point is recorded as overflow opportunity Rud; at the current measuring point, the median value of the overflow opportunity in the adjacent area of ​​any second characteristic unit is defined as the regional overflow opportunity, and other second characteristic units that are closest to the current second characteristic unit and whose overflow opportunity is greater than the regional overflow opportunity of the current second characteristic unit are searched, and the distance between the two is recorded as the overflow opportunity radius; a circle formed by taking any second characteristic unit as the center and the overflow opportunity radius as the radius is taken as the overflow opportunity area, and the average distance from each first characteristic unit in the overflow opportunity area to the current second characteristic unit is calculated, and recorded as the inter-element analytical coefficient Inc; The intrusion analysis value Trso is calculated based on the balanced overflow phase value. If the grid unit is the first characteristic unit, the intrusion analysis value is not calculated; if the grid unit does not belong to the second characteristic unit at the current measuring point, the overflow opportunity area when it was the second characteristic unit most recently is used; the most recent time as the second characteristic unit refers to the first measuring point in the reverse time direction where the grid unit belongs to the second characteristic unit, and the intrusion analysis value is calculated based on the measuring point of the second characteristic unit.

4. The method for identifying mangrove ecological region invasion areas based on hyperspectral images according to claim 1 is characterized in that: The time series feature quantity includes one or more of an autoregressive coefficient, a moving average coefficient, an ADF test statistic, a sample entropy, an FFT component, a linear trend slope, an autocorrelation coefficient, and a partial autocorrelation coefficient.

5. The method for identifying mangrove ecological region invasion areas based on hyperspectral images according to claim 1 is characterized in that: In step S600, the method for identifying the types of invasive species in the mangrove ecological zone in combination with the invasion determination results of each invasive species is: obtaining the invasive species for which each invasion determination result is true as the determined invasive species, and the types of invasive species in the mangrove ecological zone in any grid unit include all the determined invasive species obtained in the grid unit.

6. The invasive species identification system in mangrove ecological areas based on hyperspectral images is characterized by: The invasive species identification system for mangrove ecological areas based on hyperspectral images includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the steps in the invasive area identification method for mangrove ecological areas based on hyperspectral images described in any one of claims 1 to 5 are implemented. The invasive species identification system for mangrove ecological areas based on hyperspectral images runs on desktop computers, laptop computers, PDAs, and computing devices in cloud data centers.

Images