Long-time-sequence spartina alterniflora distribution area identification method based on GEE and multi-source remote sensing data

Abstract

The invention discloses a long-time-sequence spartina alterniflora distribution area identification method based on GEE and multi-source remote sensing data, and the method comprises the following steps: A, obtaining at least two series of satellite remote sensing images of an identification area, and enabling each series of satellite remote sensing images to form a remote sensing data set; b, carrying out modularization processing on all remote sensing data sets; and C, performing image fusion processing on the remote sensing data subjected to modularization processing, and identifying a spartina alterniflora distribution area and spatial and temporal changes thereof. The method can improve the defects in the prior art, improves the recognition precision and recognition speed, and facilitates the recognition of the characteristics of the distribution area at the initial stage of field planting of spartina alterniflora.

Description

Technical Field

[0001] The invention relates to the technical field of environmental remote sensing monitoring, in particular to a method for identifying a long-time-series Spartina alterniflora distribution area based on GEE and multi-source remote sensing data. Background Art

[0002] Spartina alterniflora is an alien plant that was originally used to protect tidal flats and riverbanks. However, due to its strong reproductive capacity and lack of natural enemies, it has grown rapidly in my country's coastal wetlands in recent years, causing serious impacts on the ecosystem. In order to effectively monitor Spartina alterniflora, satellite remote sensing technology has been widely used. However, existing remote sensing data processing and recognition methods cannot take into account both speed and accuracy in the recognition process of plant distribution areas such as Spartina alterniflora. Summary of the invention

[0003] The technical problem to be solved by the present invention is to provide a long-time series Spartina alterniflora distribution area identification method based on GEE and multi-source remote sensing data, which can solve the shortcomings of the existing technology, improve the recognition accuracy and recognition speed, and is conducive to identifying the distribution area characteristics of Spartina alterniflora in the early stage of colonization.

[0004] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows.

[0005] A method for identifying the distribution area of ​​Spartina alterniflora based on long-time series of GEE and multi-source remote sensing data, comprising the following steps:

[0006] A. Obtain at least two series of satellite remote sensing images of the identified area, and each series of satellite remote sensing images constitutes a remote sensing dataset;

[0007] B. Modularize all remote sensing data sets;

[0008] C. Perform image fusion processing on the modularized remote sensing data to identify the distribution area of ​​Spartina alterniflora and its temporal and spatial changes.

[0009] Preferably, multi-period 10-meter-resolution satellite remote sensing images of Sentinal-2 of the identified area are obtained from the GEE platform, and multi-period 10-meter-resolution satellite remote sensing images of SPOT of the identified area are obtained from the French Space Research Center CNES.

[0010] Preferably, in step B, firstly, all remote sensing data sets are subjected to cloud removal processing, including the following steps:

[0011] Generate neural network models for the two remote sensing datasets, SPOT and SENTINAL-2, respectively, use the historical data of the corresponding remote sensing dataset to generate a training dataset for each neural network model, and use the training dataset to train the corresponding neural network model;

[0012] The trained neural network model is used to divide the images of the remote sensing data set into blocks, and a corresponding weight value is assigned to each image block. The image blocks are weighted and recombined to obtain the remote sensing data after de-clouding.

[0013] Preferably, a bidirectional transfer function pair is arranged between the hidden layers of the two neural network models, the bidirectional transfer function pair includes a penalty term, and the bidirectional transfer function pair is used to share the features extracted by the hidden layers of the two neural network models. After the hidden layers obtain new features, they are combined with the current features.

[0014] Preferably, the modular processing of the SENTINAL-2 satellite remote sensing data set comprises the following steps:

[0015] According to different spectral images in the SENTINAL-2 satellite remote sensing dataset, the suitable growth area of ​​Spartina alterniflora was divided, and features were extracted from other areas in the remote sensing image. Blurred processing was performed while retaining the features, and then the suitable growth area of ​​Spartina alterniflora and other areas after fuzzy processing were recombined into a remote sensing image.

[0016] Preferably, for other areas in the remote sensing image, firstly, Gaussian filtering is performed for denoising, then the image area is traversed, the grayscale gradient vector of each pixel is calculated, and while calculating, it is determined whether the grayscale gradient size of the current pixel is the local maximum value in its neighborhood in the direction of the grayscale gradient of the pixel, if so, the pixel is retained for the next round of comparison, otherwise the pixel is no longer used for comparison; all local maximum pixels are combined into a feature set, the feature set is linearized, and a linear feature set of other areas in the remote sensing image is obtained;

[0017] The fuzzy processing for other areas in the remote sensing image includes the following steps: first, a fuzzy window matrix is ​​defined, the elements in the fuzzy window matrix are weight values ​​of corresponding pixels, the fuzzy window matrix is ​​used to traverse the area to be processed, and weighted processing is performed. When the fuzzy window matrix includes a pixel of a linear feature set, the pixel is not weighted, and the grayscale mean of the pixel and its adjacent pixels is calculated at the same time, and then the grayscale mean is used to replace the grayscale of the pixel adjacent to the linear feature set pixel;

[0018] Before the remote sensing image is reassembled, the edge of the reassembled image block is feathered, and after the remote sensing image is reassembled, the position of the image splicing is sharpened.

[0019] Preferably, the modular processing of the SPOT satellite remote sensing data set comprises the following steps:

[0020] The modularized SENTINAL-2 satellite remote sensing images are arranged in time sequence, and the SPOT satellite remote sensing images are matched with the SENTINAL-2 satellite remote sensing images that are closest in time sequence. The linear feature sets in the matching SPOT satellite remote sensing images are marked on the SENTINAL-2 satellite remote sensing images, and the position transfer functions of the SENTINAL-2 satellite remote sensing images and the SPOT satellite remote sensing images are calculated according to the position deviations of the same linear feature sets in the SENTINAL-2 satellite remote sensing images and the SPOT satellite remote sensing images. The position transfer function is used to mark the suitable growth area of ​​Spartina odorifera on the SPOT satellite remote sensing image in the matching SENTINAL-2 satellite remote sensing image.

[0021] Preferably, identifying the distribution area of ​​Spartina alterniflora comprises the following steps:

[0022] The normalized vegetation index of the suitable growth area of ​​Spartina alterniflora in the SENTINAL-2 satellite remote sensing image was calculated, and then the normalized vegetation index was used to fit the growth curve of Spartina alterniflora in each suitable growth area of ​​Spartina alterniflora.

[0023] Read the full-band reflectance data of the suitable growth area of ​​Spartina alterniflora in the SPOT satellite remote sensing image, and mark the distribution area of ​​Spartina alterniflora in the SPOT satellite remote sensing image according to the reflectance characteristics of Spartina alterniflora in each band;

[0024] The growth curve of S. alterniflora was corrected using the distribution area of ​​S. alterniflora in the SPOT satellite remote sensing images according to the time series, and the distribution area of ​​S. alterniflora was inverted and calculated using the corrected growth curve. Then the spatiotemporal changes of the distribution area of ​​S. alterniflora were obtained using the changes in the distribution area of ​​S. alterniflora in the time series.

[0025] The beneficial effect brought about by adopting the above technical solution is that the present invention comprehensively utilizes two groups of satellite remote sensing images, SPOT and SENTINAL-2, and reduces the amount of calculation without affecting the recognition accuracy by optimizing the processing process of remote sensing images, thereby improving the recognition speed of remote sensing images. DETAILED DESCRIPTION

[0026] The existing technology for processing and identifying remote sensing images usually processes the remote sensing images in a unified process, and then awakens the object recognition based on the processing results. Since the monitoring range of Spartina alterniflora is very wide, the number of remote sensing images involved is huge, which leads to the current stage of processing and identifying remote sensing images of Spartina alterniflora that consumes a lot of computing power and time.

[0027] To address this problem, we abandoned the traditional remote sensing image processing method and redesigned a set of remote sensing image processing and recognition algorithms based on GEE platform resources according to the growth and geographical distribution characteristics of Spartina alterniflora. The specific steps include the following steps.

[0028] A. Obtain multi-period 10-meter-resolution satellite remote sensing images of Sentinal-2 of the identified area from the GEE platform, and obtain multi-period 10-meter-resolution satellite remote sensing images of SPOT of the identified area from the French Space Research Center CNES. Each series of satellite remote sensing images constitutes a remote sensing dataset.

[0029] B. First, all remote sensing data sets are de-clouded, including the following steps:

[0030] Generate neural network models for two remote sensing datasets, SPOT and SENTINAL-2, respectively. A bidirectional transfer function pair is set between the hidden layers of the two neural network models. The bidirectional transfer function pair includes a penalty term. The bidirectional transfer function pair is used to share the features extracted by the hidden layers of the two neural network models. After the hidden layers obtain new features, they are combined with the current features. Use the historical data of the corresponding remote sensing dataset to generate a training dataset for each neural network model, and use the training dataset to train the corresponding neural network model;

[0031] The trained neural network model is used to divide the images of the remote sensing data set into blocks, and a corresponding weight value is assigned to each image block. The image blocks are weighted and recombined to obtain the remote sensing data after de-clouding.

[0032] The modular processing of SENTINAL-2 satellite remote sensing datasets includes the following steps:

[0033] According to different spectral images in the SENTINAL-2 satellite remote sensing dataset, the suitable growth area of ​​Spartina alterniflora was divided, and features were extracted from other areas in the remote sensing image. Blurred processing was performed while retaining the features, and then the suitable growth area of ​​Spartina alterniflora and other areas after fuzzy processing were recombined into a remote sensing image.

[0034] For other areas in the remote sensing image, Gaussian filtering is first performed for denoising, and then the image area is traversed to calculate the grayscale gradient vector of each pixel. While calculating, it is determined whether the grayscale gradient size of the current pixel is the local maximum in its neighborhood in the direction of the pixel's grayscale gradient. If so, the pixel is retained for the next round of comparison, otherwise the pixel is no longer used for comparison; all local maximum pixels are combined into a feature set, and the feature set is linearized to obtain a linear feature set for other areas in the remote sensing image.

[0035] Fuzzy processing for other areas in remote sensing images includes the following steps: first, a fuzzy window matrix is ​​defined, and the elements in the fuzzy window matrix are the weight values ​​of the corresponding pixels. The fuzzy window matrix is ​​used to traverse the area to be processed and perform weighted processing. When the fuzzy window matrix includes pixels of a linear feature set, the pixel is not weighted. At the same time, the grayscale mean of the pixel and its adjacent pixels is calculated, and then the grayscale mean is used to replace the grayscale of the pixels adjacent to the linear feature set pixels.

[0036] Before the remote sensing image is reassembled, the edge of the reassembled image block is feathered, and after the remote sensing image is reassembled, the position of the image splicing is sharpened.

[0037] The modular processing of SPOT satellite remote sensing datasets includes the following steps:

[0038] The modularized SENTINAL-2 satellite remote sensing images are arranged in time sequence, and the SPOT satellite remote sensing images are matched with the SENTINAL-2 satellite remote sensing images that are closest in time sequence. The linear feature sets in the matching SPOT satellite remote sensing images are marked on the SENTINAL-2 satellite remote sensing images, and the position transfer functions of the SENTINAL-2 satellite remote sensing images and the SPOT satellite remote sensing images are calculated according to the position deviations of the same linear feature sets in the SENTINAL-2 satellite remote sensing images and the SPOT satellite remote sensing images. The position transfer function is used to mark the suitable growth area of ​​Spartina odorifera on the SPOT satellite remote sensing image in the matching SENTINAL-2 satellite remote sensing image.

[0039] C. Perform image fusion processing on the modularized remote sensing data to identify the distribution area of ​​Spartina alterniflora and its temporal and spatial changes. The identification process includes the following steps:

[0040] The normalized vegetation index of the suitable growth area of ​​Spartina alterniflora in the SENTINAL-2 satellite remote sensing image was calculated, and then the normalized vegetation index was used to fit the growth curve of Spartina alterniflora in each suitable growth area of ​​Spartina alterniflora.

[0041] Read the full-band reflectance data of the suitable growth area of ​​Spartina alterniflora in the SPOT satellite remote sensing image, and mark the distribution area of ​​Spartina alterniflora in the SPOT satellite remote sensing image according to the reflectance characteristics of Spartina alterniflora in each band;

[0042] The growth curve of S. alterniflora was corrected using the distribution area of ​​S. alterniflora in the SPOT satellite remote sensing images according to the time series, and the distribution area of ​​S. alterniflora was inverted and calculated using the corrected growth curve. Then the spatiotemporal changes of the distribution area of ​​S. alterniflora were obtained using the changes in the distribution area of ​​S. alterniflora in the time series.

[0043] The present invention selects two representative series of satellite remote sensing images, SPOT and SENTINAL-2, for image processing and recognition. The first thing to do in the preprocessing of remote sensing images is to remove the clouds in the image. Based on the existing use of neural network models for cloud removal, the present invention uses the correspondence between the two groups of remote sensing images in time and space, and sets a bidirectional transfer function between the two neural network models to share the features extracted from the hidden layers of the two neural network models. This can not only accelerate the training convergence speed of the model and improve the F1 score, but also effectively improve the generalization ability of the model when removing clouds, thereby speeding up the image processing speed.

[0044] Traditional remote sensing image recognition methods usually use remote sensing images of different bands for direct regional recognition. Due to the large amount of remote sensing image data, this recognition method usually uses the Normalized Difference Vegetation Index (NDVI) for indirect recognition, which leads to poor recognition accuracy. Since we use two series of satellite remote sensing images for image recognition, we can use the two series of images for mutual correction to improve recognition accuracy. However, the increase in the number of images will lead to a decrease in recognition speed. In response to this contradiction, we improved the image recognition algorithm. Since the shooting cycle of the SENTINAL-2 satellite is much shorter than that of the SPOT satellite, we selected the remote sensing images of the SENTINAL-2 satellite as the main image set, and calculated the linear feature sets of the suitable growth areas of Spartina alterniflora and other areas. Then, the remote sensing images of the SPOT satellite are matched with the remote sensing images of the SENTINAL-2 satellite using the linear feature sets, and the suitable growth areas of Spartina alterniflora are quickly marked on the SPOT satellite remote sensing images. In this way, the separate processing of the SPOT satellite remote sensing images is omitted. When the distribution area of ​​Spartina alterniflora is identified, only the suitable growth area of ​​Spartina alterniflora in the SENTINAL-2 satellite remote sensing image is accurately identified to obtain the growth curve of Spartina alterniflora, and the identification area is greatly reduced, which can improve the recognition speed. Then, the growth curve of Spartina alterniflora is corrected by using the distribution area of ​​Spartina alterniflora in the SPOT satellite remote sensing image (because the number of SPOT satellite remote sensing images is small, the suitable growth area of ​​Spartina alterniflora is processed twice without bringing an obvious increase in the amount of calculation) to improve the recognition accuracy.

[0045] The present invention comprehensively utilizes two series of satellite remote sensing images, SPOT and SENTINAL-2, to identify the distribution area of ​​Spartina alterniflora in a long time series, with fast speed and high precision, providing a new way for environmental monitoring of coastal wetlands. Moreover, the identification method provided by the present invention also has a wide range of application scenarios in other fields. Next, we will conduct in-depth research on this identification method to expand its applicable fields.

[0046] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for identifying the distribution area of ​​Spartina alterniflora based on long-term GEE and multi-source remote sensing data, characterized in that The following steps are involved: A. Obtain at least two series of satellite remote sensing images of the identified area, and each series of satellite remote sensing images constitutes a remote sensing dataset; B. Modularize all remote sensing data sets; C. Perform image fusion processing on the modularized remote sensing data to identify the distribution area of ​​Spartina alterniflora and its temporal and spatial changes.

2. The method for identifying the distribution area of ​​Spartina alterniflora based on GEE and multi-source remote sensing data in a long time series according to claim 1, characterized in that: The multi-period 10-meter-resolution satellite remote sensing images of Sentinal-2 in the identified area were obtained from the GEE platform, and the multi-period 10-meter-resolution satellite remote sensing images of SPOT in the identified area were obtained from the French Space Research Center CNES.

3. The method for identifying the distribution area of ​​Spartina alterniflora based on GEE and multi-source remote sensing data in a long time series according to claim 2, characterized in that: In step B, all remote sensing data sets are first de-clouded, including the following steps: Generate neural network models for the two remote sensing datasets, Sentinal-2 and SPOT, respectively, use the historical data of the corresponding remote sensing dataset to generate a training dataset for each neural network model, and use the training dataset to train the corresponding neural network model; The trained neural network model is used to divide the images of the remote sensing data set into blocks, and a corresponding weight value is assigned to each image block. The image blocks are weighted and recombined to obtain the remote sensing data after de-clouding.

4. The method for identifying the distribution area of ​​Spartina alterniflora based on GEE and multi-source remote sensing data in a long time series according to claim 3, characterized in that: A bidirectional transfer function pair is set between the hidden layers of the two neural network models. The bidirectional transfer function pair includes a penalty term and is used to share the features extracted by the hidden layers of the two neural network models. After the hidden layers obtain new features, they are combined with the current features.

5. The method for identifying the distribution area of ​​Spartina alterniflora based on GEE and multi-source remote sensing data in a long time series according to claim 3, characterized in that: The modular processing of the Sentinal-2 satellite remote sensing dataset consists of the following steps, According to different spectral images in the Sentinal-2 satellite remote sensing dataset, the suitable growth area of ​​Spartina alterniflora was divided, and features were extracted from other areas in the remote sensing image. Blurred processing was performed while retaining the features, and then the suitable growth area of ​​Spartina alterniflora and other areas after fuzzy processing were recombined into a remote sensing image.

6. The method for identifying the distribution area of ​​Spartina alterniflora based on GEE and multi-source remote sensing data in a long time series according to claim 5, characterized in that: For other areas in the remote sensing image, first perform Gaussian filtering for denoising, then traverse the image area, calculate the grayscale gradient vector of each pixel, and while calculating, determine whether the grayscale gradient size of the current pixel is the local maximum in its neighborhood in the direction of the pixel's grayscale gradient. If so, retain the pixel for the next round of comparison, otherwise, do not use the pixel for comparison; combine all local maximum pixels into a feature set, linearize the feature set, and obtain the linear feature set of other areas in the remote sensing image; The fuzzy processing for other areas in the remote sensing image includes the following steps: first, a fuzzy window matrix is ​​defined, the elements in the fuzzy window matrix are weight values ​​of corresponding pixels, the fuzzy window matrix is ​​used to traverse the area to be processed, and weighted processing is performed. When the fuzzy window matrix includes a pixel of a linear feature set, the pixel is not weighted, and the grayscale mean of the pixel and its adjacent pixels is calculated at the same time, and then the grayscale mean is used to replace the grayscale of the pixel adjacent to the linear feature set pixel; Before the remote sensing image is reassembled, the edge of the reassembled image block is feathered, and after the remote sensing image is reassembled, the position of the image splicing is sharpened.

7. The method for identifying the distribution area of ​​Spartina alterniflora based on GEE and multi-source remote sensing data in a long time series according to claim 6, characterized in that: The modular processing of SPOT satellite remote sensing datasets includes the following steps: The modularized SENTINAL-2 satellite remote sensing images are arranged in time sequence, and the SPOT satellite remote sensing images are matched with the SENTINAL-2 satellite remote sensing images that are closest in time sequence. The linear feature sets in the matching SPOT satellite remote sensing images are marked on the SENTINAL-2 satellite remote sensing images, and the position transfer functions of the SENTINAL-2 satellite remote sensing images and the SPOT satellite remote sensing images are calculated according to the position deviations of the same linear feature sets in the SENTINAL-2 satellite remote sensing images and the SPOT satellite remote sensing images. The position transfer function is used to mark the suitable growth area of ​​Spartina odorifera on the SPOT satellite remote sensing image in the matching SENTINAL-2 satellite remote sensing image.

8. The method for identifying the distribution area of ​​Spartina alterniflora based on long-time series GEE and multi-source remote sensing data according to claim 7, characterized in that: Identifying the distribution area of ​​Spartina alterniflora involves the following steps: The normalized vegetation index of the suitable growth area of ​​Spartina alterniflora in the SENTINAL-2 satellite remote sensing image was calculated, and then the normalized vegetation index was used to fit the growth curve of Spartina alterniflora in each suitable growth area of ​​Spartina alterniflora. Read the full-band reflectance data of the suitable growth area of ​​Spartina alterniflora in the SPOT satellite remote sensing image, and mark the distribution area of ​​Spartina alterniflora in the SPOT satellite remote sensing image according to the reflectance characteristics of Spartina alterniflora in each band; The growth curve of S. alterniflora was corrected using the distribution area of ​​S. alterniflora in the SPOT satellite remote sensing images according to the time series, and the distribution area of ​​S. alterniflora was inverted and calculated using the corrected growth curve. Then the spatiotemporal changes of the distribution area of ​​S. alterniflora were obtained using the changes in the distribution area of ​​S. alterniflora in the time series.