Seaweed, seaweed and coral distinguishing method and device based on phenological characteristics

Through spectral index analysis of remote sensing image time series and fitting curves to extract phenological characteristics, combined with RF classification method, the problem of remote sensing technology being difficult to distinguish seagrass, seaweeds and corals is solved, and the accuracy of seagrass distribution map is improved.

CN119068348APending Publication Date: 2024-12-03SUZHOU UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410924449.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing remote sensing technology is difficult to accurately distinguish seaweed, seaweed and coral, resulting in low accuracy in mapping seaweed distribution maps.

Method used

Based on the time series of remote sensing images, spectral index is calculated, and the vegetation characteristic index value set is constructed. By fitting the curve, phenological characteristics are extracted, classification characteristic space is constructed, and RF classification method is combined to achieve the distinction between seaweed, seaweed and coral.

Benefits of technology

It improves the accuracy of distinguishing seaweed, seaweed and corals, reduces the error caused by spectral feature distinction, and improves the accuracy of drawing seaweed distribution maps.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068348A_ABST
    Figure CN119068348A_ABST
Patent Text Reader

Abstract

The invention relates to a sea grass, seaweed and coral distinguishing method and device based on phenological characteristics, and the method comprises the steps: obtaining a remote sensing image time sequence of a to-be-recognized region containing sea grass, seaweed and coral based on preset sampling time, calculating the spectral index of each remote sensing image, constructing a vegetation characteristic index value set, and obtaining a vegetation characteristic index value set; performing pixel-level time sequence reconstruction to obtain a sea grass feature fitting curve, a seaweed feature fitting curve and a coral feature fitting curve, respectively extracting various phenological features of sea grass, seaweed and coral, and constructing a sea grass classification feature space, a seaweed classification feature space and a coral classification feature space; performing feature cascading with preset image features of the remote sensing image time sequence to obtain a sea grass fusion feature information base, a seaweed fusion feature information base and a coral fusion feature information base, and classifying each remote sensing image in the remote sensing image time sequence by using an RF classification method to obtain a classification result; and sea grass, seaweed and coral in the to-be-identified area are distinguished.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing images, and in particular to a method and device for distinguishing seaweed, seaweed and coral based on phenological characteristics. Background Art

[0002] Seagrass is a higher aquatic plant that grows in shallow ocean waters. It has high ecological value and is a key component of the marine blue carbon ecosystem. However, seagrass has been declining rapidly around the world in recent years, so it is of great significance to protect seagrass and its ecological functions.

[0003] Seagrass surveys are the prerequisite for implementing seagrass protection. The current conventional seagrass survey method is based on field observations, which has the disadvantages of long operation cycles, high implementation costs, and difficulty in detecting unknown waters. Therefore, remote sensing technology has begun to be used to survey seagrass. Satellite remote sensing technology not only has the ability to make large-scale statistical inferences quickly and easily, but can also quickly, conveniently, and efficiently achieve large-scale seagrass distribution mapping at a lower cost. It has significant advantages over traditional field observation survey methods, but the current use of remote sensing technology to survey seagrass has the problem of low accuracy.

[0004] In addition to seagrass, there are two common biological communities, seaweed and coral, in the shallow water near the coast of the ocean. Since algae contain chlorophyll, and corals have zooxanthellae that also contain chlorophyll, the spectra of seagrass and coral are similar. Therefore, in the process of using remote sensing technology to investigate seagrass, the existing methods only rely on the spectral characteristics of the ground objects of remote sensing technology to distinguish, and it is difficult to accurately distinguish seagrass from algae and coral, resulting in low accuracy in seagrass investigation and low accuracy in drawing seagrass distribution maps. Summary of the invention

[0005] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the existing technology cannot accurately distinguish seagrass, seaweed and coral by only using the spectral characteristics of the objects using remote sensing technology, resulting in low accuracy in drawing seagrass distribution maps.

[0006] In order to solve the above technical problems, the present invention provides a method for distinguishing seaweed, seaweed and coral based on phenological characteristics, comprising:

[0007] Based on the preset sampling time, a time series of remote sensing images of the area to be identified, including seagrass, seaweed and coral, is obtained;

[0008] Calculating the spectral index of each remote sensing image in the remote sensing image time series to construct a vegetation characteristic index value set;

[0009] Reconstructing the vegetation characteristic index value set at the pixel level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve, and a coral characteristic fitting curve;

[0010] Based on the inflection points, slopes and amplitudes of the seagrass characteristic fitting curves, seaweed characteristic fitting curves and coral characteristic fitting curves, various phenological characteristics of seagrass, seaweed and coral were extracted respectively, and the seagrass classification feature space, seaweed classification feature space and coral classification feature space were constructed;

[0011] The seagrass classification feature space, the seaweed classification feature space and the coral classification feature space are respectively cascaded with the preset image features of the remote sensing image time series to obtain a seagrass fusion feature information library, a seaweed fusion feature information library and a coral fusion feature information library;

[0012] Based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database, the RF classification method is used to classify each remote sensing image in the remote sensing image time series to distinguish between seagrass, seaweed and coral in the area to be identified.

[0013] Preferably, the spectral index of each remote sensing image in the remote sensing image time series is calculated, which is expressed as:

[0014] Spectral Index

[0015] Among them, NIR represents the near infrared band in the remote sensing image, and R represents the red light band in the remote sensing image.

[0016] Preferably, the multiple phenological characteristics include the start time of the growing season, the end time of the growing season, a preset benchmark value, a seasonal amplitude, a growth rate at the beginning of the season and a decrease rate at the end of the season.

[0017] Preferably, the pixel-level time series reconstruction of the vegetation characteristic index value set to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve and a coral characteristic fitting curve comprises:

[0018] The spectral index in the vegetation characteristic index value set is smoothed by using the SAM method, and multiple segments of seaweed time-series growth curves, multiple segments of seaweed time-series growth curves and multiple segments of coral time-series growth curves are obtained based on the preset time length;

[0019] The double Logistic fitting function was used to fit the multi-segment seagrass time-series growth curves, the multi-segment seaweed time-series growth curves and the multi-segment coral time-series growth curves to obtain the seagrass characteristic fitting curves, the seaweed characteristic fitting curves and the coral characteristic fitting curves.

[0020] Preferably, a dual Logistic fitting function is used to adopt a piecewise local fitting method to obtain a characteristic fitting curve based on multiple segments of time series growth curves.

[0021] Preferably, the characteristic fitting curve is expressed as:

[0022] f(t)=f(t;c1,c2,m1,m2,n1,n2)=c1+c2g(t;m1,m2,n1,n2);

[0023]

[0024] Among them, t represents the tth remote sensing image, c1 represents the baseline value of the characteristic fitting curve, c2 represents the seasonal amplitude; m1 is the first inflection point of the characteristic fitting curve, indicating the start time of the growing season; m2 is the second inflection point of the characteristic fitting curve, indicating the end time of the growing season; n1 is the rate of change of the first inflection point of the characteristic fitting curve, indicating the growth rate at the beginning of the season; n2 is the rate of change of the second inflection point of the characteristic fitting curve, indicating the decreasing rate at the end of the season.

[0025] Preferably, the preset image features of the remote sensing image time series include texture features, shape structure features and spatial relationship features.

[0026] Preferably, the method of classifying each remote sensing image in the remote sensing image time series by using the RF classification method based on the seagrass fusion feature information library, the seaweed fusion feature information library and the coral fusion feature information library to achieve the distinction between seagrass, seaweed and coral in the area to be identified includes:

[0027] The seaweed fusion features, seaweed fusion features and coral fusion features in the seaweed fusion feature information library, the seaweed fusion feature information library and the coral fusion feature information library are respectively used as inputs of the RF model, and the category labels corresponding to the fusion features are used as outputs, and the RF model is trained until a preset number of training times is reached to obtain a target RF model; the category labels include seaweed, seaweed and coral;

[0028] Input each remote sensing image in the remote sensing image time series into the target RF model to obtain the category label corresponding to each pixel in each remote sensing image;

[0029] Obtain the category label corresponding to each pixel in all remote sensing images in the remote sensing image time series to complete the distinction between seagrass, seaweed and coral in the area to be identified.

[0030] Preferably, before calculating the spectral index NDVI of each remote sensing image in the remote sensing image time series, each remote sensing image is also preprocessed; the preprocessing includes radiation calibration, atmospheric correction, geometric correction, registration and stitching and cropping.

[0031] This embodiment provides a device for distinguishing seaweed, seaweed and coral based on phenological characteristics, including:

[0032] The vegetation characteristic index value set construction module is used to obtain a remote sensing image time series of the area to be identified containing seagrass, seaweed and coral based on a preset sampling time; calculate the spectral index of each remote sensing image in the remote sensing image time series to construct a vegetation characteristic index value set;

[0033] A curve fitting module is used to reconstruct the vegetation characteristic index value set at the pixel level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve and a coral characteristic fitting curve;

[0034] The classification feature space construction module is used to extract various phenological characteristics of seagrass, seaweed and coral based on the inflection points, slopes and amplitudes of the seagrass feature fitting curves, seaweed feature fitting curves and coral feature fitting curves, and to construct the seagrass classification feature space, seaweed classification feature space and coral classification feature space;

[0035] A feature information library construction module is used to perform feature cascade on the seagrass classification feature space, the seaweed classification feature space and the coral classification feature space with the preset image features of the remote sensing image time series, respectively, to obtain a seagrass fusion feature information library, a seaweed fusion feature information library and a coral fusion feature information library;

[0036] The RF classification module is used to classify each remote sensing image in the remote sensing image time series based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database, using the RF classification method to distinguish seagrass, seaweed and coral in the area to be identified.

[0037] The above technical solution of the present invention has the following beneficial effects compared with the prior art:

[0038] The method for distinguishing seaweed, seaweed and coral based on phenological characteristics of the present invention takes into account the different phenological characteristics of the growth cycles of seaweed, seaweed and coral, that is, the curve law of chlorophyll content changing with time is different. Therefore, the growth curve is fitted based on the remote sensing image time series of the area to be identified, and the phenological characteristics of seaweed, seaweed and coral are extracted based on the fitted growth curve; the phenological characteristics are cascaded with the preset image characteristics to obtain the corresponding feature information library to train the RF model, so as to realize the classification of the three biological communities of seaweed, seaweed and coral. The present invention makes a preliminary distinction between seaweed, seaweed and coral based on the remote sensing image time series, thereby extracting the corresponding phenological characteristics therefrom, and then combining the phenological characteristics with the remote sensing image characteristics to obtain more accurate fusion characteristics, and training the RF classification model based on the fusion characteristics, thereby improving the model classification accuracy, reducing the error caused by distinguishing seaweed, seaweed and coral only by relying on spectral characteristics, improving the accuracy of seaweed identification and monitoring, and improving the accuracy of seaweed distribution map drawing. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein

[0040] Figure 1 It is a flowchart of the steps of the method for distinguishing seaweed, seaweed and coral based on phenological characteristics provided by the present invention;

[0041] Figure 2 is the seaweed characteristic fitting curve constructed by the present invention;

[0042] Figure 3 It is a detailed flow chart of the method for distinguishing seaweed, seaweed and coral provided by the present invention. DETAILED DESCRIPTION

[0043] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0044] In shallow sea ecosystems, seagrass, algae and corals are often distributed in an interlaced manner, and their spectral characteristics are also similar. Therefore, it is difficult to distinguish the three during the image interpretation process, which reduces the accuracy of seagrass remote sensing monitoring. However, the phenological cycles of seagrass, algae and corals during their growth are different: as a higher plant community in the ocean, seagrass shows significant periodic changes in growth in spring and summer and withering in autumn and winter, so the chlorophyll concentration is high in summer and low in winter. The chlorophyll concentration in corals is closely negatively correlated with seawater temperature and photosynthetically active radiation, showing the characteristics of low in summer and high in winter. The growth cycle of algae is also inconsistent with that of seagrass, and often has a certain lag or advance. Because they have significant differences in growth cycle fluctuations, the spectral information of seagrass, algae and corals on remote sensing images also has significant periodic differences. Therefore, the present invention will effectively distinguish seagrass from algae and corals by obtaining the differences in the phenological characteristics of seagrass and other marine organisms in time series remote sensing images.

[0045] Reference Figure 1 As shown, it is a flowchart of the steps of the method for distinguishing seaweed, seaweed and coral based on phenological characteristics of the present invention, and the specific distinguishing steps include:

[0046] S101: based on a preset sampling time, obtaining a remote sensing image time series of the area to be identified including seagrass, seaweed and coral;

[0047] S102: Calculate the spectral index of each remote sensing image in the remote sensing image time series to construct a vegetation characteristic index value set;

[0048] Spectral Index

[0049] Among them, NIR represents the near infrared band in remote sensing images, and R represents the red light band in remote sensing images;

[0050] S103: reconstructing the vegetation characteristic index value set at the pixel level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve, and a coral characteristic fitting curve;

[0051] S104: Based on the inflection points, slopes and amplitudes of the seagrass characteristic fitting curves, the seaweed characteristic fitting curves and the coral characteristic fitting curves, various phenological characteristics of seagrass, seaweed and coral are extracted respectively, and the seagrass classification feature space, the seaweed classification feature space and the coral classification feature space are constructed;

[0052] S105: performing feature cascade on the seagrass classification feature space, the seaweed classification feature space, and the coral classification feature space with the preset image features of the remote sensing image time series, respectively, to obtain a seagrass fusion feature information library, a seaweed fusion feature information library, and a coral fusion feature information library;

[0053] The preset image features of the remote sensing image time series include texture features, shape structure features and spatial relationship features;

[0054] S106: Based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database, each remote sensing image in the remote sensing image time series is classified using the RF classification method to distinguish seagrass, seaweed and coral in the area to be identified.

[0055] Specifically, in step S103, the vegetation characteristic index value set is reconstructed in pixel-level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve, and a coral characteristic fitting curve, including:

[0056] S103-1: Smoothing the spectral index in the vegetation characteristic index value set by using the SAM method, and obtaining multiple seagrass time series growth curves, multiple seaweed time series growth curves, and multiple coral time series growth curves based on a preset time length;

[0057] S103-2: Using a double Logistic fitting function, the multiple segments of seagrass time series growth curves, the multiple segments of seaweed time series growth curves and the multiple segments of coral time series growth curves are fitted respectively to obtain the seagrass characteristic fitting curve, the seaweed characteristic fitting curve and the coral characteristic fitting curve.

[0058] Among them, the double Logistic fitting function is used to adopt the piecewise local fitting method, and the characteristic fitting curve is obtained based on the multi-segment time series growth curve; the characteristic fitting curve is expressed as:

[0059] f(t)=f(t;c1,c2,m1,m2,n1,n2)=c1+c2g(t;m1,m2,n1,n2);

[0060]

[0061] Among them, t represents the tth remote sensing image, c1 represents the baseline value of the characteristic fitting curve, c2 represents the seasonal amplitude; m1 is the first inflection point of the characteristic fitting curve, indicating the start time of the growing season; m2 is the second inflection point of the characteristic fitting curve, indicating the end time of the growing season; n1 is the rate of change of the first inflection point of the characteristic fitting curve, indicating the growth rate at the beginning of the season; n2 is the rate of change of the second inflection point of the characteristic fitting curve, indicating the decreasing rate at the end of the season.

[0062] Reference Figure 2 As shown in the figure, it is a fitting curve of seaweed characteristics constructed by an embodiment of the present invention; in this embodiment, the phenological characteristic variables can be roughly known from the smoothed growth curve, and the specific parameters of the phenological characteristic variables can be accurately known from the fitted growth curve. Specifically, the SMA method is first used to smooth the original NDVI growth curve and eliminate noise, and then the Figure 2 Such a curve is obtained, and six phenological characteristic variables are obtained from the known time series cycle; then the double Logistic method is used to perform curve fitting operation; according to Figure 2 The phenological characteristic information parameters of the vegetation growth cycle are obtained from the characteristic point positions shown in . Therefore, the corresponding phenological characteristics can be obtained based on the characteristic fitting curve, and the various phenological characteristics include the start time of the growing season, the end time of the growing season, the preset reference value, the seasonal amplitude, the growth rate at the beginning of the season, and the decline rate at the end of the season.

[0063] In this embodiment, the smoothing of the growth curve is to remove abnormal points and reduce errors for a growth curve of a known time period, the purpose of which is to eliminate errors and express the most basic fluctuation law of the growth curve; the fitting of the curve is to predict the changing trend of the curve in the future for a longer period of time based on the fluctuation law of the known growth curve, the purpose of which is to be able to obtain the fluctuation law of the curve in a longer period of time and to highlight the phenological characteristics of growth from multiple change periods. In other words, the purpose of constructing the fitting curve is to obtain a vegetation growth curve with a longer period of time, more accurate and complete, so that the phenological characteristic variable parameters of vegetation growth can be more easily captured.

[0064] The embodiment of the present invention obtains the time series NDVI value from the remote sensing image time series, constructs the time series vegetation characteristic index set, and obtains the NDVI growth curve within a certain time period. However, this curve has noise in reflecting the vegetation growth cycle. In order to eliminate the influence of noise and reduce errors, the time series NDVI value within a certain time period should be optimized, so the SMA method is used to smooth the curve. When the smooth growth curve within a certain time period is obtained, in order to obtain a growth curve with a longer time period and to highlight the phenological characteristics, it is necessary to use the double Logistic method to fit the smooth growth curve to obtain the final fitting curve, that is, the smoothed fitting curve.

[0065] Specifically, in step S106, based on the seagrass fusion feature information library, the seaweed fusion feature information library and the coral fusion feature information library, the RF classification method is used to classify each remote sensing image in the remote sensing image time series to distinguish seagrass, seaweed and coral in the area to be identified, including:

[0066] S106-1: using the seaweed fusion features, the seaweed fusion features and the coral fusion features in the seaweed fusion feature information library, the seaweed fusion feature information library and the coral fusion feature information library as inputs of the RF model, and using the category labels corresponding to the fusion features as outputs, training the RF model until a preset number of training times is reached, and obtaining a target RF model; the category labels include seaweed, seaweed and coral;

[0067] S106-2: Input each remote sensing image in the remote sensing image time series into the target RF model to obtain the category label corresponding to each pixel in each remote sensing image;

[0068] S106-3: Obtain the category label corresponding to each pixel in all remote sensing images in the remote sensing image time series, and complete the distinction between seagrass, seaweed and coral in the area to be identified.

[0069] After acquiring the remote sensing image time series, this embodiment also includes preprocessing each remote sensing image before calculating the spectral index NDVI of each remote sensing image; the preprocessing includes radiation calibration, atmospheric correction, geometric correction, registration, and stitching and cropping. Radiation calibration allows data from different sensors to be compared and analyzed on the same physical scale; atmospheric correction aims to eliminate the influence of the atmosphere on the reflected or emitted radiation from the surface, thereby improving the quality of remote sensing images; geometric correction is used to correct geometric distortion in remote sensing images to ensure that the location of objects in the image is consistent with the actual geographical location; registration is used to accurately align remote sensing images acquired at different times, different sensors or different angles in geographic space, ensuring that the locations of objects between different images can correspond one to one, facilitating operations such as change detection and time series analysis; stitching is used to merge multiple adjacent remote sensing images into a larger seamless image, and cropping is the process of extracting the region of interest ROI from the stitched image or the original image to reduce the complexity of data processing and the need for storage space. Through these preprocessing steps, the quality of remote sensing images can be significantly improved, providing an accurate and reliable data basis for calculating the spectral index NDVI.

[0070] Based on the above embodiment, since the phenological characteristics of the growth cycles of seagrass, seaweed and coral are different, that is, the curves of chlorophyll content changing over time are different, this difference can be used to extract the phenological characteristics of the three on remote sensing images, and the phenological characteristics can be integrated into the classification process of remote sensing images to extract image information of the three biological communities of seagrass, seaweed and coral. In the embodiment of the present invention, the seagrass, seaweed and coral distinguishing method provided by the present invention is used to distinguish seagrass, seaweed and coral in the monitoring area, and the difference is referred to as Figure 3 The following is a detailed flow chart of the method for distinguishing seaweed, seaweed and coral. The specific steps for distinguishing include:

[0071] S201: Obtain satellite image data of the monitoring area for each month within 5 years and construct a time series original remote sensing image dataset;

[0072] S202: Preprocessing of each phase of image data, including radiometric calibration, atmospheric correction, geometric correction, image registration, stitching and cropping;

[0073] S203: Calculate NDVI (Normalized Difference Vegetation Index) for each period of preprocessed image data to construct a time series remote sensing image feature data set;

[0074] The calculation formula of NDVI is:

[0075] S204: For the time series NDVI image characteristic index dataset, the smooth moving average (SMA) method is used to determine the seagrass growing season, including six phenological characteristic variables such as the start time of the growing season, the end time of the growing season, the baseline value, the seasonal amplitude, the growth rate at the beginning of the season, and the decline rate at the end of the season;

[0076] S205: For the time series NDVI image characteristic index data set, a double Logistic function is used to fit the time series NDVI, and the noise effect in the data is eliminated by smoothing the time series curve to improve the quality of the time series characteristic index image;

[0077] S206: For the improved time series NDVI image set, identify the differences in phenological rules of algae, seaweed and coral, and extract the phenological characteristic information of the three;

[0078] S207: Cascading the extracted phenological characteristic information of seagrass, seaweed and coral with other image characteristic information to construct an image characteristic information database of seagrass, seaweed and coral; other characteristics refer to common image characteristic information, including texture characteristics, shape structure characteristics, spatial relationship characteristics, etc.;

[0079] S208: Based on the phenological characteristic information of seagrass, seaweed and coral and other image feature information databases, the original remote sensing image data are classified using the Random Forest RF (Random Forest) classification algorithm;

[0080] S209: Conduct field verification and accuracy evaluation of the classification results of seagrass, seaweed and coral; complete the differentiation of seagrass, seaweed and coral based on time series remote sensing images and phenological characteristics, and map the distribution of the three.

[0081] The present invention adopts a classical fitting method to reconstruct a pixel-level time series for a vegetation characteristic index set, so as to eliminate noise information in the time series curve and extract phenological features. The time series curve is fitted with a function form that matches the vegetation phenological growth law to obtain a smoother fitting curve that is more in line with the vegetation growth phenological law. Phenological features are extracted from the fitting curve and cascaded with image features to distinguish seagrass, seaweed and coral based on the cascaded features. The present embodiment can effectively distinguish seagrass from seaweed and coral, reduce the error caused by seaweed and coral caused by spectral similarity in seagrass remote sensing monitoring, and improve the accuracy of seagrass identification and monitoring.

[0082] Based on the above embodiments, an embodiment of the present invention provides a device for distinguishing seaweed, seaweed and coral based on phenological characteristics, and the specific device includes:

[0083] The vegetation characteristic index value set construction module is used to obtain a remote sensing image time series of the area to be identified containing seagrass, seaweed and coral based on a preset sampling time; calculate the spectral index of each remote sensing image in the remote sensing image time series to construct a vegetation characteristic index value set;

[0084] A curve fitting module is used to reconstruct the vegetation characteristic index value set at the pixel level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve and a coral characteristic fitting curve;

[0085] The classification feature space construction module is used to extract various phenological characteristics of seagrass, seaweed and coral based on the inflection points, slopes and amplitudes of the seagrass feature fitting curves, seaweed feature fitting curves and coral feature fitting curves, and to construct the seagrass classification feature space, seaweed classification feature space and coral classification feature space;

[0086] A feature information library construction module is used to perform feature cascade on the seagrass classification feature space, the seaweed classification feature space and the coral classification feature space with the preset image features of the remote sensing image time series, respectively, to obtain a seagrass fusion feature information library, a seaweed fusion feature information library and a coral fusion feature information library;

[0087] The RF classification module is used to classify each remote sensing image in the remote sensing image time series based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database, using the RF classification method to distinguish seagrass, seaweed and coral in the area to be identified.

[0088] The device for distinguishing seaweed, seaweed and coral based on phenological characteristics of this embodiment is used to implement the aforementioned method for distinguishing seaweed, seaweed and coral based on phenological characteristics. Therefore, the specific implementation methods of the device for distinguishing seaweed, seaweed and coral based on phenological characteristics can be seen in the embodiment part of the method for distinguishing seaweed, seaweed and coral based on phenological characteristics in the previous text. For example, the vegetation characteristic index value set construction module 100 is used to implement steps S101 and S102 in the above-mentioned method for distinguishing seaweed, seaweed and coral based on phenological characteristics; the curve fitting module 200, the classification feature space construction module 300, the feature information library construction module 400, and the RF classification module 500 are respectively used to implement steps S103, S104, S105 and S106 in the above-mentioned method for distinguishing seaweed, seaweed and coral based on phenological characteristics. Therefore, its specific implementation methods can refer to the description of the corresponding embodiments of each part, which will not be repeated here.

[0089] The method and device for distinguishing seaweed, seaweed and coral based on phenological characteristics of the present invention take into account the different phenological characteristics of the growth cycles of seaweed, seaweed and coral, that is, the curve law of chlorophyll content changing with time is different. Therefore, the growth curve is fitted based on the remote sensing image time series of the area to be identified, and the phenological characteristics of seaweed, seaweed and coral are extracted based on the fitted growth curve; the phenological characteristics are cascaded with the preset image characteristics to obtain the corresponding feature information library to train the RF model, so as to realize the classification of the three biological communities of seaweed, seaweed and coral. The present invention makes a preliminary distinction between seaweed, seaweed and coral based on the remote sensing image time series, thereby extracting the corresponding phenological characteristics therefrom, and then combining the phenological characteristics with the remote sensing image characteristics to obtain more accurate fusion characteristics, and training the RF classification model based on the fusion characteristics, thereby improving the classification accuracy of the model, reducing the error caused by distinguishing seaweed, seaweed and coral only by relying on spectral characteristics, improving the accuracy of seaweed identification and monitoring, and improving the accuracy of seaweed distribution map drawing.

[0090] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0091] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0092] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0093] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0094] Obviously, the above embodiments are merely examples for the purpose of clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the present invention.

Claims

1. A method for distinguishing seagrass, seaweed and coral based on phenological characteristics. It is characterized in that include: Based on the preset sampling time, a time series of remote sensing images of the area to be identified, including seagrass, seaweed and coral, is obtained; Calculating the spectral index of each remote sensing image in the remote sensing image time series to construct a vegetation characteristic index value set; Reconstructing the vegetation characteristic index value set at the pixel level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve, and a coral characteristic fitting curve; Based on the inflection points, slopes and amplitudes of the seagrass characteristic fitting curves, seaweed characteristic fitting curves and coral characteristic fitting curves, various phenological characteristics of seagrass, seaweed and coral were extracted respectively, and the seagrass classification feature space, seaweed classification feature space and coral classification feature space were constructed; The seagrass classification feature space, the seaweed classification feature space and the coral classification feature space are respectively cascaded with the preset image features of the remote sensing image time series to obtain a seagrass fusion feature information library, a seaweed fusion feature information library and a coral fusion feature information library; Based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database, the RF classification method is used to classify each remote sensing image in the remote sensing image time series to distinguish between seagrass, seaweed and coral in the area to be identified.

2. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 1, It is characterized in that The spectral index of each remote sensing image in the remote sensing image time series is calculated, which is expressed as: Spectral Index Among them, NIR represents the near infrared band in the remote sensing image, and R represents the red light band in the remote sensing image.

3. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 1, It is characterized in that The various phenological characteristics include the start time of the growing season, the end time of the growing season, a preset benchmark value, a seasonal amplitude, a growth rate at the beginning of the season and a decrease rate at the end of the season.

4. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 1, It is characterized in that The pixel-level time series reconstruction of the vegetation characteristic index value set to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve, and a coral characteristic fitting curve includes: The spectral index in the vegetation characteristic index value set is smoothed by using the SAM method, and multiple segments of seaweed time-series growth curves, multiple segments of seaweed time-series growth curves and multiple segments of coral time-series growth curves are obtained based on the preset time length; The double logistic fitting function was used to fit the multi-segment seagrass time-series growth curves, the multi-segment seaweed time-series growth curves and the multi-segment coral time-series growth curves to obtain the seagrass characteristic fitting curves, the seaweed characteristic fitting curves and the coral characteristic fitting curves.

5. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 4, It is characterized in that The double logistic fitting function was used to adopt the piecewise local fitting method and obtain the characteristic fitting curve based on multiple time series growth curves.

6. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 5, It is characterized in that The characteristic fitting curve is expressed as: f(t)=f(t;c 1 ,c 2 ,m 1 ,m 2 ,n 1 ,n 2 )=c 1 +c 2 g(t;m 1 ,m 2 ,n 1 ,n 2 ); Among them, t represents the tth remote sensing image, c 1 Indicates the reference value of the characteristic fitting curve, c 2 represents the seasonal amplitude; m 1 is the first inflection point of the characteristic fitting curve, indicating the start time of the growing season; m 2 is the second inflection point of the characteristic fitting curve, indicating the end of the growing season; n 1 is the rate of change of the first inflection point of the characteristic fitting curve, indicating the growth rate at the beginning of the quarter; n 2 It is the rate of change of the second inflection point of the characteristic fitting curve, indicating the end-of-season decline rate.

7. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 1, It is characterized in that The preset image features of the remote sensing image time series include texture features, shape structure features and spatial relationship features.

8. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 1, It is characterized in that The method uses the RF classification method based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database to classify each remote sensing image in the remote sensing image time series, so as to distinguish seagrass, seaweed and coral in the area to be identified, including: The seaweed fusion features, seaweed fusion features and coral fusion features in the seaweed fusion feature information library, the seaweed fusion feature information library and the coral fusion feature information library are respectively used as inputs of the RF model, and the category labels corresponding to the fusion features are used as outputs, and the RF model is trained until a preset number of training times is reached to obtain a target RF model; the category labels include seaweed, seaweed and coral; Input each remote sensing image in the remote sensing image time series into the target RF model to obtain the category label corresponding to each pixel in each remote sensing image; Obtain the category label corresponding to each pixel in all remote sensing images in the remote sensing image time series to complete the distinction between seagrass, seaweed and coral in the area to be identified.

9. The method for distinguishing seaweed, seaweed and coral based on phenological characteristics according to claim 1, It is characterized in that Before calculating the spectral index NDVI of each remote sensing image in the remote sensing image time series, each remote sensing image is preprocessed; the preprocessing includes radiation calibration, atmospheric correction, geometric correction, registration and splicing and cropping.

10. A device for distinguishing seaweed, seaweed and coral based on phenological characteristics. It is characterized in that include: The vegetation characteristic index value set construction module is used to obtain a remote sensing image time series of the area to be identified containing seagrass, seaweed and coral based on a preset sampling time; calculate the spectral index of each remote sensing image in the remote sensing image time series to construct a vegetation characteristic index value set; A curve fitting module is used to reconstruct the vegetation characteristic index value set at the pixel level time series to obtain a seagrass characteristic fitting curve, a seaweed characteristic fitting curve and a coral characteristic fitting curve; The classification feature space construction module is used to extract various phenological characteristics of seagrass, seaweed and coral based on the inflection points, slopes and amplitudes of the seagrass feature fitting curves, seaweed feature fitting curves and coral feature fitting curves, and to construct the seagrass classification feature space, seaweed classification feature space and coral classification feature space; A feature information library construction module is used to perform feature cascade on the seagrass classification feature space, the seaweed classification feature space and the coral classification feature space with the preset image features of the remote sensing image time series, respectively, to obtain a seagrass fusion feature information library, a seaweed fusion feature information library and a coral fusion feature information library; The RF classification module is used to classify each remote sensing image in the remote sensing image time series based on the seagrass fusion feature information database, the seaweed fusion feature information database and the coral fusion feature information database, using the RF classification method to distinguish seagrass, seaweed and coral in the area to be identified.