Laver cultivation area extraction method and device, readable storage medium and electronic equipment
By combining optical and radar remote sensing image features, and utilizing random forest classifiers and morphological processing, the problems of classification accuracy and cross-year monitoring in seaweed farming areas were solved, achieving high-precision monitoring of seaweed farming areas around the clock.
Patent Information
- Application Number
- CN202211086108.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-06
- Publication Date
- 2026-01-16
- Estimated Expiration
- 2042-09-06
AI Technical Summary
Traditional classification methods for seaweed farming areas based on optical remote sensing data are not very accurate under the influence of clouds and fog, and lack methods for cross-year migration, resulting in high monitoring costs and low efficiency.
By combining optical remote sensing images and radar image sequences, target spectral features, polarization features, multispectral index features and radar index features are extracted. Random forest classifier is used to classify and extract laver farming areas, and morphological processing is used to improve accuracy.
It enables the classification and extraction of laver from laver farming areas around the clock and in all weather conditions, improving accuracy and possessing regional portability and cross-year application capabilities, making it suitable for laver farming areas with large-scale and complex structures.
Smart Images
Figure CN115661471B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present disclosure relates to the fields of computer technology, remote sensing technology, ecology, etc., in particular, relates to a porphyra farming area extraction method and device, readable storage medium and electronic equipment. BACKGROUND
[0002] Porphyra dianoe as a large seaweed rich in nutrients such as protein, vitamins, minerals and dietary fiber, is an important economic cultivated seaweed in the coastal zone. In recent years, scholars have conducted a lot of research on the monitoring of seaweed cultivation. Traditional survey-based methods consume a lot of manpower, material and financial resources, and cannot meet the grasp of spatial distribution. In contrast, remote sensing technology has the advantages of low cost, near real-time, wide monitoring range, etc. At present, it is mostly based on optical remote sensing data, and machine learning methods are used to extract porphyra farming areas. However, due to the influence of optical remote sensing data by clouds and fog, the accuracy of the classification and extraction results of the porphyra farming area cannot be guaranteed. At the same time, due to the limitation of different years of light intensity, precipitation and other factors, there is still a lack of classification methods that can be used across years. SUMMARY
[0003] In order to overcome the problems in the related art, the present disclosure provides a porphyra farming area extraction method, device, readable storage medium and electronic equipment.
[0004] In order to achieve the above purpose, in a first aspect, the present disclosure provides a porphyra farming area extraction method, comprising:
[0005] Obtaining target remote sensing image data of a target area, wherein the target remote sensing image data includes a first target multispectral image and a target radar image sequence, and the acquisition time of the first target multispectral image is within the acquisition time period of the target radar image sequence;
[0006] Extracting target spectral features, target polarization features, target multispectral index features and target radar index features of the target remote sensing image data;
[0007] According to the target spectral features, the target polarization features, the target multispectral index features and the target radar index features, the porphyra farming area in the target area is classified and extracted.
[0008] Optionally, the extraction of the target spectral features, the target polarization features, the target multispectral index features and the target radar index features of the target remote sensing image data comprises:
[0009] Performing atmospheric correction, cloud mask processing and resampling processing on the first target multispectral image in turn to obtain a second target multispectral image;
[0010] synthesizing the target radar image sequence to obtain a median image;
[0011] determining target sea-land mask data of the target region in a target year, wherein the target year is a year in which the first target multi-spectral image is collected;
[0012] generating target spectral features, target polarization features, target multi-spectral index features and target radar index features of the target remote sensing image data according to the second target multi-spectral image, the median image and the target sea-land mask data.
[0013] Optionally, the target multi-spectral index features include target spectral water body index features and target vegetation index features, and the target radar index features include target radar water body index features.
[0014] The generating of the target spectral features, the target polarization features, the target multi-spectral index features and the target radar index features of the target remote sensing image data according to the second target multi-spectral image, the median image and the target sea-land mask data includes:
[0015] performing continent region elimination on the second target multi-spectral image and the median image respectively according to the target sea-land mask data;
[0016] extracting the target spectral features from the second target multi-spectral image obtained after the elimination, and generating the target vegetation index features and the target spectral water body index features according to the target spectral features;
[0017] extracting the target polarization features from the median image obtained after the elimination;
[0018] generating the target radar water body index features according to the target polarization features.
[0019] Optionally, the determining of the target sea-land mask data of the target region in the target year includes:
[0020] obtaining a first multi-spectral image sequence of the target region in the target year;
[0021] performing cloud screening on the first multi-spectral image sequence to obtain a second multi-spectral image sequence;
[0022] performing cloud mask processing on each historical multi-spectral image in the second multi-spectral image sequence;
[0023] synthesizing all the historical multi-spectral images obtained after the cloud mask processing by using a maximum spectral index synthesis method to obtain a maximum water surface image of the target region;
[0024] According to the maximum water surface image, the target region is segmented into sea and land by using the Otsu method to obtain target sea and land mask data of the target region in the target year.
[0025] Optionally, the classification and extraction of the laver cultivation area in the target region according to the target spectral feature, the target polarization feature, the target multispectral index feature, and the target radar index feature comprises:
[0026] The target spectral feature, the target polarization feature, the target multispectral index feature, and the target radar index feature are input into a pre-trained random forest classifier to classify and extract the laver cultivation area in the target region.
[0027] Optionally, the random forest classifier is trained in the following manner:
[0028] Sample remote sensing image data of a sample region and a classification sample set of a laver cultivation area in the sample remote sensing image data are obtained, wherein the sample remote sensing image data comprises a first sample multispectral image and a sample radar image sequence, and the acquisition time of the first sample multispectral image is within the acquisition time period of the sample radar image sequence.
[0029] Sample spectral features, sample polarization features, sample multispectral index features, and sample radar index features of the sample remote sensing image data are extracted.
[0030] The sample spectral features, the sample polarization features, the sample multispectral index features, and the sample radar index features are used as inputs of the random forest classifier, and the classification sample set of the laver cultivation area in the sample remote sensing image data is used as the target output of the random forest classifier to perform model training in this manner, so as to obtain the random forest classifier.
[0031] Optionally, the method further comprises:
[0032] The classification and extraction result of the laver cultivation area in the target region is subjected to morphological processing, wherein the morphological processing comprises at least one of clustering, hole filling, edge smoothing, and isolated point filtering.
[0033] In a second aspect, the present disclosure provides a laver cultivation area extraction device, comprising:
[0034] A first acquisition module is configured to acquire target remote sensing image data of a target region, wherein the target remote sensing image data comprises a first target multispectral image and a target radar image sequence, and the acquisition time of the first target multispectral image is within the acquisition time period of the target radar image sequence.
[0035] a first feature extraction module configured to extract target spectral features, target polarization features, target multispectral index features, and target radar index features of the target remote sensing image data acquired by the first acquisition module;
[0036] a classification extraction module configured to classify and extract the seaweed farming area in the target region based on the target spectral features, the target polarization features, the target multispectral index features, and the target radar index features extracted by the first feature extraction module.
[0037] In a third aspect, the present disclosure provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the seaweed farming area extraction method provided in the first aspect of the present disclosure.
[0038] In a fourth aspect, the present disclosure provides an electronic device, comprising:
[0039] a memory having a computer program stored thereon;
[0040] a processor configured to execute the computer program in the memory to implement the steps of the seaweed farming area extraction method provided in the first aspect of the present disclosure.
[0041] In the above technical solution, when classifying and extracting the seaweed farming area in the target region, not only optical remote sensing images but also radar image sequences are referred to, so that the characteristics of radar microwaves can penetrate clouds and fog and are the only data that can be used in extreme weather conditions, which makes the above-mentioned seaweed farming area extraction method not affected by the climate or weather, and has the characteristics of all-day and all-weather application. Moreover, since the radar image data has high temporal and spatial resolution and is not easily affected by clouds and fog, the classification and extraction accuracy of the seaweed farming area can be improved. In addition, when classifying and extracting the seaweed farming area in the target region, the spectral features, polarization features, target multispectral index features, and target radar index features of the remote sensing image data are simultaneously referred to, which can make the features used for classification and extraction of the seaweed farming area more abundant, and further improve the classification and extraction accuracy of the seaweed farming area. In addition, the above-mentioned seaweed farming area extraction method has regional portability and can be used for classification and extraction of any large-scale seaweed farming area with complex structure and wide area, and can be used across years.
[0042] Other features and advantages of the present disclosure will be described in detail in the following detailed description section. BRIEF DESCRIPTION OF DRAWINGS
[0043] The accompanying drawings are included to provide a further understanding of the present disclosure and constitute a part of the specification, illustrate the present disclosure and are used to explain the present disclosure together with the specific embodiments described below, but do not constitute a limitation on the present disclosure. In the drawings:
[0044] Figure 1 is a flowchart of a method for extracting a seaweed cultivation area according to an exemplary embodiment.
[0045] Figure 2 is a flowchart of a method for training a random forest classifier according to an exemplary embodiment.
[0046] Figure 3 is a flowchart of a method for extracting a seaweed cultivation area according to another exemplary embodiment.
[0047] Figure 4 is a block diagram of an apparatus for extracting a seaweed cultivation area according to an exemplary embodiment.
[0048] Figure 5 is a block diagram of an electronic device according to an exemplary embodiment.
[0049] Figure 6 is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION
[0050] The specific embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely intended to illustrate and explain the present disclosure, and are not intended to limit the present disclosure.
[0051] Figure 1 is a flowchart of a method for extracting a seaweed cultivation area according to an exemplary embodiment. The method can be applied to a terminal or a server, for example, a Google Earth Engine (GEE) platform, which is the most widely used remote sensing cloud computing platform. Since the GEE is a cloud computing platform supporting a large amount of geospatial data, the method for extracting a seaweed cultivation area and batch processing of remote sensing images become simple, and the cost and complexity of classification and extraction of a seaweed cultivation area can be reduced. Therefore, the method for extracting a seaweed cultivation area is preferably applied to the GEE platform. As shown in Figure 1 The method for extracting a seaweed cultivation area can include the following S101-S103.
[0052] In S101, target remote sensing image data of a target area is acquired.
[0053] In the present disclosure, the target region can be any region to be classified and extracted for laver cultivation. The target remote sensing image data includes a first target multispectral image and a target radar image sequence, wherein the acquisition time of the first target multispectral image is within the acquisition time period of the target radar image sequence. For example, the acquisition time period of the target radar image sequence is the month in which the acquisition time of the first target multispectral image is located.
[0054] The first target multispectral image collected by the multispectral instrument in the remote sensing satellite and the target radar image sequence collected by the synthetic aperture radar in the remote sensing satellite can be directly received, or historical remote sensing data obtained by the remote sensing satellite and stored in the server side (for example, the GEE platform) can be used. The target remote sensing data of the target region can be directly obtained through remote communication, for example, through 2G network, 3G network, 4G network or 5G network, wireless fidelity (WIFI) network, etc. for communication, or indirectly obtained through U disk, mobile hard disk, etc. That is, the present embodiment does not limit the way of obtaining the target remote sensing data of the target region.
[0055] In S102, the target spectral feature, the target polarization feature, the target multispectral index feature and the target radar index feature of the target remote sensing image data are extracted.
[0056] In the present disclosure, the target spectral feature can include at least one of visible light band data, near-infrared band data and short-infrared band data in the target remote sensing image, wherein the visible light band data can include red, green and blue band data.
[0057] The target polarization feature can include vertical-vertical (VV) polarization data and vertical-horizontal (VH) polarization data.
[0058] The target multispectral index feature includes a target spectral water body index feature and a target vegetation index feature, and the target radar index feature includes a target radar water body index feature. The target spectral water body index can be a modified normalized difference water body index (MNDWI), and the target radar water body index can be a Sentinel-1 dual-polarized water body index. The target vegetation index feature can include at least one of an enhanced index feature, a vegetation difference index feature and a normalized vegetation index feature.
[0059] In S103, the seaweed cultivation area in the target region is classified and extracted according to the target spectral feature, the target polarization feature, the target multispectral index feature, and the target radar index feature.
[0060] In the above technical solution, when the seaweed cultivation area in the target region is classified and extracted, not only the optical remote sensing image is referred to, but also the radar image sequence is referred to. In this way, the characteristics that the radar microwaves can penetrate clouds and fog and are the only data that can be used in extreme weather conditions are fully utilized, so that the above seaweed cultivation area extraction method is not affected by the climate or weather, and has the characteristics of all-day and all-weather application. Moreover, since the radar image data has high temporal and spatial resolution and is not easily affected by clouds and fog, the classification and extraction accuracy of the seaweed cultivation area can be improved. In addition, when the seaweed cultivation area in the target region is classified and extracted, the spectral feature, polarization feature, and index feature of the remote sensing image data are simultaneously referred to, which can make the features used for the classification and extraction of the seaweed cultivation area more abundant, and further improve the classification and extraction accuracy of the seaweed cultivation area. In addition, the above seaweed cultivation area extraction method has regional portability and can be used for the classification and extraction of any large-scale seaweed cultivation area with complex structure and wide area, and can be used across years.
[0061] The specific implementation of the target spectral feature, the target polarization feature, the target multispectral index feature, and the target radar index feature of the target remote sensing image data in S102 will be described in detail below. Specifically, the following steps (1) to (4) can be used to achieve this:
[0062] (1) The first target multispectral image is sequentially subjected to atmospheric correction, cloud mask processing, and resampling processing to obtain a second target multispectral image.
[0063] (2) The target radar image sequence is synthesized to obtain a median image.
[0064] (3) The target sea-land mask data of the target region in the target year is determined.
[0065] The target year is the year in which the acquisition time of the first target multispectral image is located. For example, if the acquisition time of the first target multispectral image is May 1, 2022, the target year is 2022, i.e., the target sea-land mask data is the sea-land mask data of the target region in 2022.
[0066] (4) The target spectral feature, the target polarization feature, the target multispectral index feature, and the target radar index feature of the target remote sensing image data are generated according to the second target multispectral image, the median image, and the target sea-land mask data.
[0067] The specific implementation of determining the target sea-land mask data of the target region in the target year will be described in detail below.
[0068] Specifically, the determination can be implemented in various ways. In one embodiment, if the sea-land mask data of the target region in the target year exists, it is directly determined as the target sea-land mask data.
[0069] In another embodiment, the target sea-land mask data of the target region in the target year can be determined by the following steps (31) to (35):
[0070] (31) Obtain a first multi-spectral image sequence of the target region in the target year.
[0071] (32) Perform cloud screening on the first multi-spectral image sequence to obtain a second multi-spectral image sequence.
[0072] (33) For each historical multi-spectral image in the second multi-spectral image sequence, perform cloud mask processing on the historical multi-spectral image.
[0073] In the present disclosure, the cloud mask processing is performed on each historical multi-spectral image in the second multi-spectral image sequence, which can remove the influence of a large amount of thick clouds, thereby improving the accuracy of the target sea-land mask data and further improving the classification and extraction accuracy of the subsequent seaweed cultivation area.
[0074] (34) Use the maximum spectral index synthesis method to synthesize all the historical multi-spectral images obtained after cloud mask processing to obtain a maximum water surface image of the target region.
[0075] The maximum water surface image obtains the maximum coverage range of the water surface of the target region in the target year.
[0076] (35) According to the maximum water surface image, perform sea-land segmentation on the target region by using the Otsu method to obtain the target sea-land mask data of the target region in the target year.
[0077] In the present disclosure, the Otsu method (also known as the maximum inter-class variance method) is applied to automatically perform sea-land binary segmentation on the MNDWI band of the maximum water surface image to obtain the target sea-land mask data of the target region in the target year.
[0078] In the above embodiment, the target sea-land mask data of the target region in the target year is determined according to the first multi-spectral image sequence of the target region in the target year, rather than using the existing sea-land mask data, which can avoid the change of the mask data caused by land reclamation, thereby ensuring the accuracy of the target sea-land mask data and further improving the classification and extraction accuracy of the subsequent seaweed cultivation area.
[0079] The following will be described in detail with respect to the specific implementation of generating the target spectral feature, the target polarization feature, the target multispectral index feature and the target radar index feature of the target remote sensing image data according to the second target multispectral image, the median image and the target land-sea mask data in (4) above. Specifically, the following steps (41) to (44) can be implemented:
[0080] (41) According to the target land-sea mask data, the continental region is removed from the second target multispectral image and the median image respectively.
[0081] (42) The target spectral feature is extracted from the second target multispectral image obtained after removal, and the target vegetation index feature and the target spectral water body index feature are generated according to the target spectral feature.
[0082] (43) The target polarization feature is extracted from the median image obtained after removal.
[0083] (44) The target radar water body index feature is generated according to the target polarization feature.
[0084] The following will be described in detail with respect to the specific implementation of classifying and extracting the nori culture area in the target region according to the target spectral feature, the target polarization feature, the target multispectral index feature and the target radar index feature in S103 above. Specifically, the target spectral feature, the target polarization feature, the target multispectral index feature and the target radar index feature can be input into the pre-trained random forest classifier to classify and extract the nori culture area in the target region. The random forest classifier is more robust and can ensure the reliability of the classification and extraction result.
[0085] The random forest classifier can be obtained by training S201 to S203 shown in the following table: Figure 2
[0086] In S201, the sample remote sensing image data of the sample region and the classification sample set of the nori culture area in the sample remote sensing image data are obtained.
[0087] In the present disclosure, the sample region can be selected as a region with a long coastline and rich mariculture, for example, Jiangsu Province; the sample remote sensing image data can be selected as medium-high resolution images in the nori growth period in years with less cloud cover and good data availability. The sample remote sensing image data includes a first sample multispectral image and a sample radar image sequence, and the acquisition time of the first sample multispectral image is within the acquisition time period of the sample radar image sequence.
[0088] The growth period of the laver is determined according to the phenological characteristics and the breeding law. The raft type laver breeding facility is generally placed in the sea from September to October every year, and the breeding activity is completed from the middle and late March to the early May of the next year. Therefore, the growth period of the laver is from December to the early March of the next year, that is, the period from the laver seedling to the laver picking.
[0089] In addition, the vector sample with the classification mark of the laver breeding area can be manually interpreted from the sample remote sensing image data combined with other remote sensing images (for example, Landsat historical images, Google Earth historical images), and the vector sample is converted into a target format, that is, a classification sample set of the laver breeding area in the sample remote sensing data image is obtained. The visual interpretation of the remote sensing image is to combine the image features (hue or color, that is, spectral features) and spatial features (shape, size, shadow, texture, pattern, position and layout) of the image with various non-remote sensing information data, use the biogeological correlation law to identify various target ground objects, and perform qualitative and quantitative analysis to obtain various ground information required.
[0090] The target format can be, for example, a JPG, PNG, TIFF raster format or the like. Since the TIFF raster format can retain geographic coordinates, projection and the like, preferably, the target format is a TIFF raster format.
[0091] In S202, the sample spectral features, sample polarization features, sample multispectral index features and sample radar index features of the sample remote sensing image data are extracted.
[0092] In the present disclosure, the sample remote sensing image data includes a plurality of sample points, and the sample spectral features, sample polarization features, sample multispectral index features and sample radar index features of each sample point can be extracted respectively.
[0093] The sample spectral features can specifically include near-infrared band data in the sample remote sensing image, and can also include visible light band data and / or short-wave infrared band data.
[0094] The sample polarization features can include VV polarization data and VH polarization data.
[0095] The sample multispectral index features include sample spectral water body index features and sample vegetation index features, and the sample radar index features include sample radar water body index features. The sample spectral water body index can be an MNDWI, and the sample radar water body index can be a Sentinel-1 dual-polarization water body index. The sample vegetation index features can include at least one of an enhanced index feature, a vegetation difference index feature, and a normalized vegetation index feature.
[0096] In S203, a model is trained by taking the sample spectral features, the sample polarization features, the sample multispectral index features, and the sample radar index features as inputs of a random forest classifier, and taking a classification sample set of the seaweed cultivation area in the sample remote sensing image data as a target output of the random forest classifier, to obtain the random forest classifier.
[0097] In addition, when the target region is an intertidal zone, the intertidal zone is affected by tides and may have soil exposed, so the target multispectral index features can include a target soil index feature (which can be determined according to the target spectral features) in addition to the target spectral water body index features and the target vegetation index features, thereby further improving the classification and extraction accuracy of the seaweed cultivation area. Accordingly, the sample index features can include sample soil index features in addition to the sample water body index features and the sample vegetation index features.
[0098] In addition, to further improve the classification and extraction accuracy of the seaweed cultivation area, morphological processing can be performed on the classification and extraction result of the seaweed cultivation area in the target region after the classification and extraction of the seaweed cultivation area in the target region, which can avoid pixel fragmentation of the classification and extraction result, thereby further improving the classification and extraction accuracy of the seaweed cultivation area. Specifically, as shown in Figure 3 The seaweed cultivation area extraction method can further include S104 as shown in
[0099] In S104, morphological processing is performed on the classification and extraction result of the seaweed cultivation area in the target region.
[0100] In the present disclosure, the morphological processing can include at least one of clustering, hole filling, edge smoothing, and isolated point filtering. The clustering is to cluster and merge adjacent similar segmentation regions using mathematical morphology operators (erosion and dilation); most of the spots will disappear in the generated classification and extraction result by filling the holes, and the jagged edge information will be alleviated by edge smoothing processing. In addition, the eight-connected method can be used to filter out isolated pixels in the classification and extraction result.
[0101] Figure 4 is a block diagram of a seaweed cultivation area extraction device according to an example embodiment. As shown in Figure 4As shown, the apparatus 400 can include:
[0102] a first acquisition module 401 configured to acquire target remote sensing image data of a target region, wherein the target remote sensing image data includes a first target multispectral image and a target radar image sequence, and a collection time of the first target multispectral image is located within a collection time period of the target radar image sequence;
[0103] a first feature extraction module 402 configured to extract target spectral features, target polarization features, target multispectral index features, and target radar index features of the target remote sensing image data acquired by the first acquisition module 401;
[0104] a classification extraction module 403 configured to perform classification extraction on a seaweed cultivation area in the target region according to the target spectral features, the target polarization features, the target multispectral index features, and the target radar index features extracted by the first feature extraction module 402.
[0105] In the above technical solution, when performing classification extraction on the seaweed cultivation area in the target region, not only optical remote sensing images are referred to, but also radar image sequences are referred to, so that the characteristics that radar microwaves can penetrate clouds and fog and are the only data that can be used in extreme weather conditions are fully utilized, so that the above seaweed cultivation area extraction method is not affected by the climate or weather and has the characteristics of being applicable at all times and in all weather. Moreover, since the radar image data has high temporal resolution and spatial resolution and is not easily affected by clouds and fog, the classification extraction accuracy of the seaweed cultivation area can be improved. In addition, when performing classification extraction on the seaweed cultivation area in the target region, the spectral features, polarization features, and index features of the remote sensing image data are simultaneously referred to, so that the features used for classification extraction of the seaweed cultivation area are more abundant, and the classification extraction accuracy of the seaweed cultivation area is further improved. In addition, the above seaweed cultivation area extraction method has regional portability and can be used for classification extraction of any large-scale seaweed cultivation area with complex structure and wide area, and can be used across years.
[0106] Optionally, the first feature extraction module 402 includes:
[0107] a first processing sub-module configured to perform, in sequence, atmospheric correction, cloud mask processing, and resampling processing on the first target multispectral image to obtain a second target multispectral image;
[0108] a first synthesis sub-module configured to synthesize the target radar image sequence to obtain a median image;
[0109] determining a target land-sea mask data of the target region in a target year, wherein the target year is a year in which the first target multi-spectral image is captured;
[0110] generating a target spectral feature, a target polarization feature, a target multi-spectral index feature and a target radar index feature of the target remote sensing image data according to the second target multi-spectral image, the median image and the target land-sea mask data.
[0111] Optionally, the target multi-spectral index feature comprises a target spectral water body index feature and a target vegetation index feature, and the target radar index feature comprises a target radar water body index feature.
[0112] The first generating sub-module comprises:
[0113] performing continent region elimination on the second target multi-spectral image and the median image respectively according to the target land-sea mask data;
[0114] generating the target vegetation index feature and the target spectral water body index feature according to the target spectral feature;
[0115] extracting the target polarization feature from the median image after elimination;
[0116] generating the target radar water body index feature according to the target polarization feature.
[0117] Optionally, the determining sub-module comprises:
[0118] acquiring a first multi-spectral image sequence of the target region in the target year;
[0119] performing cloud screening on the first multi-spectral image sequence to obtain a second multi-spectral image sequence;
[0120] performing cloud mask processing on each historical multi-spectral image in the second multi-spectral image sequence;
[0121] synthesizing all the historical multi-spectral images after cloud mask processing by using a maximum spectral index synthesis method to obtain a maximum water surface image of the target region;
[0122] The segmentation submodule is configured to perform sea-land segmentation on the target region by using the Otsu method based on the maximum water surface image, to obtain target sea-land mask data of the target region in the target year.
[0123] Optionally, the classification extraction module 403 is configured to input the target spectral feature, the target polarization feature, the target multispectral index feature, and the target radar index feature into a pre-trained random forest classifier, to classify and extract the nori cultivation area in the target region.
[0124] Optionally, the random forest classifier is obtained by model training, and the model training device comprises:
[0125] The second acquisition module is configured to acquire sample remote sensing image data of a sample region and a classification sample set of a nori cultivation area in the sample remote sensing image data, wherein the sample remote sensing image data comprises a first sample multispectral image and a sample radar image sequence, and the acquisition time of the first sample multispectral image is within the acquisition time period of the sample radar image sequence.
[0126] The second feature extraction module is configured to extract a sample spectral feature, a sample polarization feature, a sample multispectral index feature, and a sample radar index feature of the sample remote sensing image data.
[0127] The training module is configured to perform model training by taking the sample spectral feature, the sample polarization feature, the sample multispectral index feature, and the sample radar index feature as the input of the random forest classifier, and taking the classification sample set of the nori cultivation area in the sample remote sensing image data as the target output of the random forest classifier, to obtain the random forest classifier.
[0128] Optionally, the device 400 further comprises:
[0129] The processing module is configured to perform morphological processing on the classification extraction result of the nori cultivation area in the target region, wherein the morphological processing comprises at least one of clustering, hole filling, edge smoothing, and isolated point filtering.
[0130] As to the device in the above embodiments, the specific manners in which various modules perform operations have been described in detail in the embodiments of the method, and thus will not be described here in detail.
[0131] The present disclosure also provides a non-transitory computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the steps of the above nori cultivation area extraction method provided by the present disclosure.
[0132] Figure 5is a block diagram of an electronic device 700 according to an exemplary embodiment. As shown, the electronic device 700 can include a first processor 701, a first memory 702. The electronic device 700 can also include one or more of a multimedia component 703, a first input / output (I / O) interface 704, and a first communication component 705. Figure 5
[0133] The first processor 701 is configured to control overall operations of the electronic device 700 to complete all or part of the steps of the above-mentioned method for extracting a seaweed cultivation area. The first memory 702 is configured to store various types of data to support operations of the electronic device 700, which can include, for example, instructions for operating any application or method on the electronic device 700, and application-related data, such as contact data, sent and received messages, pictures, audio, video, and the like. The first memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read-only memory (EEPROM), an erasable programmable read-only memory (EPROM), a programmable read-only memory (PROM), a read-only memory (ROM), a magnetic memory, a flash memory, a magnetic disk, or an optical disk. The multimedia component 703 can include a screen and an audio component. The screen can be, for example, a touch screen, and the audio component is configured to output and / or input audio signals. For example, the audio component can include a microphone configured to receive external audio signals. The received audio signals can be further stored in the first memory 702 or transmitted through the first communication component 705. The audio component also includes at least one speaker configured to output audio signals. The first I / O interface 704 provides an interface between the first processor 701 and other interface modules, which can be a keyboard, a mouse, a button, and the like. The buttons can be virtual buttons or physical buttons. The first communication component 705 is configured to perform wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, near field communication (NFC), 2G, 3G, 4G, NB-IOT, eMTC, or other 5G, and the like, or a combination of one or more of them, is not limited herein. Therefore, the first communication component 705 can include, for example, a Wi-Fi module, a Bluetooth module, an NFC module, and the like.
[0134] In an example embodiment, the electronic device 700 can be implemented by one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), controllers, micro-controllers, microprocessors, or other electronic elements for executing the above-mentioned method of extracting a laver cultivation area.
[0135] In another example embodiment, a computer-readable storage medium including program instructions is also provided, which when executed by a processor implement the steps of the above-mentioned method of extracting a laver cultivation area. For example, the computer-readable storage medium can be the above-mentioned first memory 702 including program instructions, which can be executed by the first processor 701 of the electronic device 700 to complete the above-mentioned method of extracting a laver cultivation area.
[0136] Figure 6 is a block diagram of an electronic device 1900 according to an example embodiment. For example, the electronic device 1900 can be provided as a server. Referring to Figure 6 , the electronic device 1900 includes a second processor 1922, the number of which can be one or more, and a second memory 1932 for storing computer programs executable by the second processor 1922. The computer programs stored in the second memory 1932 can include one or more than one module each corresponding to a set of instructions. In addition, the second processor 1922 can be configured to execute the computer programs to perform the above-mentioned method of extracting a laver cultivation area.
[0137] In addition, the electronic device 1900 can further include a power supply component 1926, which can be configured to perform power management of the electronic device 1900, and a second communication component 1950, which can be configured to implement communication of the electronic device 1900, for example, wired or wireless communication. In addition, the electronic device 1900 can further include a second input / output (I / O) interface 1958. The electronic device 1900 can operate based on an operating system stored in the second memory 1932, for example, Windows Server TM , Mac OS X TM , Unix TMLinux TM and so on.
[0138] In another exemplary embodiment, a computer readable storage medium including program instructions which when executed by a processor implement the steps of the above-mentioned method for extracting a seaweed cultivation area is also provided. For example, the non-transitory computer readable storage medium can be the above-mentioned second memory 1932 including program instructions which can be executed by the second processor 1922 of the electronic device 1900 to complete the above-mentioned method for extracting a seaweed cultivation area.
[0139] In another exemplary embodiment, a computer program product containing a computer program which can be executed by a programmable apparatus, the computer program having code portions for performing the above-mentioned method for extracting a seaweed cultivation area when executed by the programmable apparatus is also provided.
[0140] The preferred embodiments of the present disclosure are described in detail above with reference to the accompanying drawings, but the present disclosure is not limited to the specific details in the above-described embodiments. Various simple modifications can be made to the technical solutions of the present disclosure within the scope of the technical concepts of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.
[0141] In addition, it should be noted that each specific technical feature described in the above specific embodiments can be combined in any appropriate manner without contradiction. In order to avoid unnecessary repetition, various possible combinations are not described again in the present disclosure.
[0142] Furthermore, any combination of the various different embodiments of the present disclosure can also be made, as long as it does not deviate from the idea of the present disclosure, it should also be considered as disclosed by the present disclosure.
Claims
1. A method for extracting a cultivation area of a porphyra, characterized by, The method comprises the following steps: acquiring target remote sensing image data of a target area, wherein the target remote sensing image data comprises a first target multispectral image and a target radar image sequence, and the acquisition time of the first target multispectral image is within the acquisition time period of the target radar image sequence; extracting target spectral features, target polarization features, target multispectral index features and target radar index features of the target remote sensing image data; classifying and extracting a seaweed cultivation area in the target area according to the target spectral features, the target polarization features, the target multispectral index features and the target radar index features; the extraction of the target spectral features, the target polarization features, the target multispectral index features and the target radar index features of the target remote sensing image data comprises: performing atmospheric correction, cloud mask processing and resampling processing on the first target multispectral image in sequence to obtain a second target multispectral image; synthesizing the target radar image sequence to obtain a median image; determining target sea-land mask data of the target area in a target year, wherein the target year is the year in which the acquisition time of the first target multispectral image is located; generating target spectral features, target polarization features, target multispectral index features and target radar index features of the target remote sensing image data according to the second target multispectral image, the median image and the target sea-land mask data.
2. The method of claim 1, wherein, The target multispectral index features comprise target spectral water body index features and target vegetation index features, and the target radar index features comprise target radar water body index features; The generation of the target spectral features, the target polarization features, the target multispectral index features and the target radar index features of the target remote sensing image data according to the second target multispectral image, the median image and the target sea-land mask data comprises: performing continent area elimination on the second target multispectral image and the median image respectively according to the target sea-land mask data; extracting the target spectral features from the second target multispectral image obtained after the elimination, and generating the target vegetation index features and the target spectral water body index features according to the target spectral features; extracting the target polarization features from the median image obtained after the elimination; generating the target radar water body index features according to the target polarization features.
3. The method of claim 1, wherein, The determination of the target sea-land mask data of the target area in the target year comprises: acquiring a first multispectral image sequence of the target area in the target year; performing cloud screening on the first multispectral image sequence to obtain a second multispectral image sequence; performing cloud mask processing on each historical multispectral image in the second multispectral image sequence; synthesizing all the historical multispectral images obtained after the cloud mask processing by using a maximum spectral index synthesis method to obtain a maximum water surface image of the target area; performing sea-land segmentation on the target area by using the Otsu method according to the maximum water surface image to obtain the target sea-land mask data of the target area in the target year.
4. The method of claim 1, wherein, The target region is classified and extracted according to the target spectral feature, the target polarization feature, the target multispectral index feature and the target radar index feature. The target region is classified and extracted according to the target spectral feature, the target polarization feature, the target multispectral index feature and the target radar index feature.
5. The method of claim 4, wherein, The random forest classifier is trained in the following manner: Sample remote sensing image data of a sample region and a classification sample set of a seaweed cultivation region in the sample remote sensing image data are obtained, wherein the sample remote sensing image data comprises a first sample multispectral image and a sample radar image sequence, and the acquisition time of the first sample multispectral image is within the acquisition time period of the sample radar image sequence; Sample spectral features, sample polarization features, sample multispectral index features and sample radar index features of the sample remote sensing image data are extracted; The random forest classifier is trained in the following manner:
6. The method according to any one of claims 1-5, characterized in that, The method further comprises: The classification extraction result of the seaweed cultivation region in the target region is subjected to morphological processing, wherein the morphological processing comprises at least one of clustering, hole filling, edge smoothing and isolated point filtering.
7. A device for extracting a cultivation area of a purple laver, characterized by comprising: The method further comprises: A first acquisition module is configured to acquire target remote sensing image data of a target region, wherein the target remote sensing image data comprises a first target multispectral image and a target radar image sequence, and the acquisition time of the first target multispectral image is within the acquisition time period of the target radar image sequence; A first feature extraction module is configured to extract target spectral features, target polarization features, target multispectral index features and target radar index features of the target remote sensing image data acquired by the first acquisition module; A classification extraction module is configured to classify and extract a seaweed cultivation region in the target region according to the target spectral features, the target polarization features, the target multispectral index features and the target radar index features extracted by the first feature extraction module; The first feature extraction module comprises: A first processing sub-module is configured to sequentially perform atmospheric correction, cloud mask processing and resampling processing on the first target multispectral image to obtain a second target multispectral image; A first synthesis sub-module is configured to synthesize the target radar image sequence to obtain a median image; A determination sub-module is configured to determine target sea-land mask data of the target region in a target year, wherein the target year is the year in which the acquisition time of the first target multispectral image is located. A first generation submodule is configured to generate target spectral features, target polarization features, target multi-spectral index features and target radar index features of the target remote sensing image data according to the second target multi-spectral image, the median image and the target sea-land mask data.
8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The program, when executed by a processor, implements the steps of the method of any one of claims 1-6.
9. An electronic device, comprising: Comprise: a memory having a computer program stored thereon; a processor configured to execute the computer program in the memory to implement the steps of the method of any one of claims 1-6.