WO2022077958A1 - 水体分割方法及装置、电子设备和存储介质 - Google Patents

水体分割方法及装置、电子设备和存储介质 Download PDF

Info

Publication number
WO2022077958A1
WO2022077958A1 PCT/CN2021/103731 CN2021103731W WO2022077958A1 WO 2022077958 A1 WO2022077958 A1 WO 2022077958A1 CN 2021103731 W CN2021103731 W CN 2021103731W WO 2022077958 A1 WO2022077958 A1 WO 2022077958A1
Authority
WO
WIPO (PCT)
Prior art keywords
water body
target image
body segmentation
segmentation result
target
Prior art date
Application number
PCT/CN2021/103731
Other languages
English (en)
French (fr)
Inventor
董润敏
Original Assignee
上海商汤智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 上海商汤智能科技有限公司 filed Critical 上海商汤智能科技有限公司
Priority to JP2022520222A priority Critical patent/JP2023502194A/ja
Publication of WO2022077958A1 publication Critical patent/WO2022077958A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10032—Satellite or aerial image; Remote sensing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning

Definitions

  • the present disclosure relates to the field of computer vision, and in particular, to a water body segmentation method and device, an electronic device and a storage medium.
  • Water body segmentation can identify and segment water body regions from remote sensing images, which is a key issue in remote sensing image processing.
  • problems in water body segmentation such as easy misdetection, often confused with buildings, roads and shadows, or it is difficult to extract narrow water bodies. Therefore, how to obtain high-quality water body segmentation results has become an urgent problem to be solved.
  • the present disclosure proposes a water body segmentation scheme.
  • a water body segmentation method comprising: segmenting a water body region in the first target image according to spectral information of the first target image to obtain a first image of the first target image. Water body segmentation result; segment the water body region in the first target image by at least one preset water body segmentation method to obtain at least one second water body segmentation result; according to the first water body of the first target image The segmentation result, and the at least one second water body segmentation result, determine the target water body segmentation result of the first target image.
  • the target water body segmentation result of the first target image is determined according to the first water body segmentation result of the first target image and the at least one second water body segmentation result , including: merging the at least one second water body segmentation result to obtain a merged water body segmentation result; combining the first water body segmentation result of the first target image and the water body area commonly included in the merged water body segmentation result , as the target water body segmentation result of the first target image.
  • the water body region in the first target image is segmented by at least one preset water body segmentation method to obtain at least one second water body segmentation result, including at least one of the following :
  • the water body data in the public map determine the water body area in the first target image, and obtain the second water body segmentation result; obtain at least one kind of surface coverage data with a resolution within a preset range, and according to the surface coverage data
  • the water body data is determined, and the water body region in the first target image is determined, and at least one second water body segmentation result is obtained.
  • the determining the water body area in the first target image according to the water body data in the public map, and obtaining the second water body segmentation result includes: according to the geographic location corresponding to the first target image range, take the water body area corresponding to the geographic range in the public map as the target water body area; if the target water body area is a water body dividing line, extend the water body dividing line in a preset direction to Preset width, take the expanded target water body area as the second water body segmentation result; in the case that the target water body area is a polygonal area and/or a circular area, take the target water body area as the second water body area Water body segmentation results.
  • the water body region in the first target image is segmented to obtain a first water body segmentation result of the first target image, including: According to the spectral information of the first target image, the normalized water index of a plurality of pixel points in the first target image is obtained; the value of the normalized water index in the plurality of pixel points is preset The pixel points within the index value range are used as the pixel points of the water body area, and the first water body segmentation result of the first target image is obtained.
  • the method further includes: according to the target water body segmentation result of the first target image, labeling the water body region in the first target image to obtain a first target image including the labeling ; Using the first target image containing the label as a sample, the initial neural network model is trained to obtain a water body segmentation network.
  • a water body segmentation method comprising: inputting a second target image into a water body segmentation network to obtain a third water body segmentation result of the second target image; information, segment the water body region of the second target image to obtain a fourth water body segmentation result of the second target image; divide the third water body segmentation result of the second target image and the fourth water body
  • the water body area commonly included in the segmentation result is used as the target water body segmentation result of the second target image.
  • the water body segmentation network is obtained by training according to the first target image and the target water body segmentation result of the first target image.
  • the method further includes: segmenting the water body region in the first target image according to the spectral information of the first target image to obtain a first image of the first target image. Water body segmentation result; segment the water body region in the first target image by at least one preset water body segmentation method to obtain at least one second water body segmentation result; according to the first water body of the first target image The segmentation result, and the at least one second water body segmentation result, determine the target water body segmentation result of the first target image.
  • the target water body segmentation result of the first target image is determined according to the first water body segmentation result of the first target image and the at least one second water body segmentation result , including: merging the at least one second water body segmentation result to obtain a merged water body segmentation result; combining the first water body segmentation result of the first target image and the water body area commonly included in the merged water body segmentation result , as the target water body segmentation result of the first target image.
  • the water body region in the first target image is segmented by at least one preset water body segmentation method to obtain at least one second water body segmentation result, including at least one of the following :
  • the water body data in the public map determine the water body area in the first target image, and obtain the second water body segmentation result; obtain at least one kind of surface coverage data with a resolution within a preset range, and according to the surface coverage data
  • the water body data is determined, and the water body region in the first target image is determined, and at least one second water body segmentation result is obtained.
  • the determining the water body area in the first target image according to the water body data in the public map, and obtaining the second water body segmentation result includes: according to the geographic location corresponding to the first target image range, take the water body area corresponding to the geographic range in the public map as the target water body area; if the target water body area is a water body dividing line, extend the water body dividing line in a preset direction to Preset width, take the expanded target water body area as the second water body segmentation result; in the case that the target water body area is a polygonal area and/or a circular area, take the target water body area as the second water body area Water body segmentation results.
  • the water body region in the first target image is segmented to obtain a first water body segmentation result of the first target image, including: According to the spectral information of the first target image, the normalized water index of a plurality of pixel points in the first target image is obtained; the value of the normalized water index in the plurality of pixel points is preset The pixel points within the index value range are used as the pixel points of the water body area, and the first water body segmentation result of the first target image is obtained.
  • the method further includes: according to the target water body segmentation result of the first target image, labeling the water body region in the first target image to obtain a first target image including the labeling ; Using the first target image containing the label as a sample, the initial neural network model is trained to obtain a water body segmentation network.
  • the water body segmentation network is trained by a preset loss function, wherein, when the water body segmentation network is trained by the preset loss function, the first training result and the third The difference between the two training results is within a preset difference range, and the first training result includes the training result obtained by taking the target water body segmentation result of the first target image as a labeled sample, and the second training result
  • the results include the training results obtained by training with the manually input water body segmentation results as the labeled samples.
  • a water body segmentation device comprising: a first water body segmentation module, configured to segment the water body region in the first target image according to the spectral information of the first target image to obtain the obtained water body region. the first water body segmentation result of the first target image; the second water body segmentation module is configured to segment the water body region in the first target image by at least one preset water body segmentation method to obtain at least one second water body segmentation result; a target water body segmentation result determination module, configured to determine the target water body segmentation of the first target image according to the first water body segmentation result of the first target image and the at least one second water body segmentation result result.
  • a water body segmentation device comprising: a third water body segmentation module for inputting a second target image into a water body segmentation network to obtain a third water body segmentation result of the second target image;
  • the four-water body segmentation module is configured to segment the water body region of the second target image according to the spectral information of the second target image to obtain a fourth water body segmentation result of the second target image; obtaining the target water body segmentation result
  • the module is configured to use the water body region commonly included in the third water body segmentation result of the second target image and the fourth water body segmentation result as the target water body segmentation result of the second target image.
  • an electronic device comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to: execute the above water body segmentation method.
  • a computer-readable storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, implement the above-mentioned method for dividing a water body.
  • the first water body segmentation result of the first target image is obtained by segmenting the water body region in the first target image according to the spectral information, and the first target image is segmented by at least one preset water body segmentation method.
  • the water body area in the image is segmented to obtain at least one second water body segmentation result, so that the target water body segmentation result of the first target image is jointly determined according to the first water body segmentation result and the at least one second water body segmentation result.
  • the first water body segmentation result determined based on the spectral information and the second water body segmentation result determined based on multiple preset water body segmentation methods can be combined, and through the mutual correction between the multiple water body segmentation results,
  • the obtained target water body segmentation results have higher quality, thereby effectively improving the accuracy and accuracy of water body segmentation.
  • FIG. 1 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 2 shows a schematic diagram of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 3 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 4 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 5 shows a schematic diagram of extending a water body dividing line according to an embodiment of the present disclosure.
  • FIG. 6 shows a schematic diagram of annotating a first target image with an expanded second water body segmentation result according to an embodiment of the present disclosure.
  • FIG. 7 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 8 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 9 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • FIG. 10 shows a block diagram of a first water body dividing apparatus according to an embodiment of the present disclosure.
  • FIG. 11 shows a block diagram of a second water body dividing apparatus according to an embodiment of the present disclosure.
  • FIG. 12 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • FIG. 13 shows a block diagram of an electronic device according to an embodiment of the present disclosure.
  • FIG. 1 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • the method can be applied to a first water body segmentation device, and the first water body segmentation device can be a terminal device, a server, or other processing devices.
  • the terminal device may be user equipment (User Equipment, UE), mobile device, user terminal, terminal, cellular phone, cordless phone, Personal Digital Assistant (PDA), handheld device, computing device, in-vehicle device, available wearable devices, etc.
  • the water body segmentation method can be applied to a cloud server or a local server, and the cloud server can be a public cloud server or a private cloud server, which can be flexibly selected according to the actual situation.
  • the water body segmentation method can also be implemented by the processor calling computer-readable instructions stored in the memory.
  • the water body segmentation method may include:
  • Step S11 segment the water body region in the first target image according to the spectral information of the first target image to obtain a first water body segmentation result of the first target image.
  • step S12 at least one preset water body segmentation method is used to segment the water body region in the first target image to obtain at least one second water body segmentation result.
  • Step S13 Determine a target water body segmentation result of the first target image according to the first water body segmentation result and at least one second water body segmentation result.
  • the first target image may be any image that has the requirement of water body segmentation, and the implementation form thereof is not limited in the embodiments of the present disclosure.
  • the first target image may be a remote sensing image, and the resolution of the remote sensing image may be flexibly determined according to the actual situation of the acquired image.
  • the first target image may be a higher resolution image. 4-band remote sensing images.
  • the source and acquisition method of the first target image are not limited in the embodiments of the present disclosure.
  • a remote sensing image directly collected by a related hardware device may be acquired. Part or all of the remote sensing images can be read or selected from the database storing the remote sensing images as the first target image.
  • the spectral information of the first target image may be spectral information of multiple bands contained in the first target image itself, and which spectral information of the bands contained in the first target image can be flexibly determined according to the actual situation of the first target image.
  • the first target image may include spectral information in the visible light band (RGB) and the near-infrared band (NIR, Near Infrared); On the basis of the spectral information of the RGB and NIR bands, more spectral information of other bands can be included.
  • step S11 since spectral information of different objects in the image may be different, the water body region in the first target image may be segmented according to the spectral information of the first target image.
  • the specific implementation manner of step S11 can be flexibly determined according to the actual situation.
  • the first water body segmentation result may include information such as the position and shape of the water body region in the first target image, and the first water body segmentation result obtained after segmentation may be displayed in an intuitive image manner, or may be stored in an information manner.
  • the manner may be flexibly determined according to the actual situation, which is not limited in the embodiments of the present disclosure.
  • step S12 by using at least one preset water body segmentation method, the implementation method of segmenting the water body region in the first target image can also be flexibly determined according to the actual situation, and reference can also be made to the following disclosed embodiments.
  • the preset water body segmentation method can be any method that can segment the water body area in the image, for example, by using a related water body segmentation calculation method, or by searching some databases containing water body segmentation results, etc., or by The neural network with the function of water body segmentation, etc., which method or methods to use can be flexibly determined according to the actual situation.
  • the water body segmentation is performed on the first target image by using several preset water body segmentation methods, and the number thereof is also not limited in the embodiment of the present disclosure.
  • the obtained second water body segmentation results may also be different. Therefore, in a possible implementation manner, the number of second water body segmentation results may correspond to the types of preset water body segmentation methods.
  • the realization form of the second water body segmentation result reference may be made to the first water body segmentation result, which will not be repeated here.
  • the realization forms of the first water body segmentation result and the second water body segmentation result may be the same or different. make restrictions.
  • first and “second” in the first water body segmentation result and the second water body segmentation result are only used to distinguish the water body segmentation results obtained by different methods, and The order of obtaining the water body segmentation results is not limited, and other numbers of the water body segmentation results in subsequent embodiments are the same, and will not be repeated.
  • step S11 and step S12 are not limited in the embodiment of the present disclosure, that is, the order of performing water body segmentation on the first target image in different ways is not limited in the embodiment of the present disclosure, in a possible implementation manner , Steps S11 and S12 may be implemented simultaneously, and in a possible implementation manner, steps S11 and S12 and the like may also be implemented respectively according to a preset implementation sequence.
  • the target water body segmentation result of the first target image may be determined according to the first water body segmentation result and the at least one second water body segmentation result.
  • the target water body segmentation result may be the final segmentation result of the water body region in the target image. Since both the first water body segmentation result and the second water body segmentation result may have inaccurate water body segmentation, in a possible implementation manner , based on the difference between the first water body segmentation result and the at least one second water body segmentation result, and adjusting on the basis of the first water body segmentation result and/or the second water body segmentation result, a more accurate target water body segmentation can be obtained result.
  • the target water body segmentation result For the realization form of the target water body segmentation result, reference may be made to the first water body segmentation result and the second water body segmentation result in the above disclosed embodiments, and details are not described herein again. How to determine the target water body segmentation result based on the first water body segmentation result and the at least one second water body segmentation result can be found in the following disclosed embodiments, which will not be described here.
  • the water body region in the first target image is segmented to obtain a first water body segmentation result of the first target image, and the first target image is segmented by at least one preset water body segmentation method. Segment the water body area in the device to obtain at least one second water body segmentation result, so that the target water body segmentation result of the first target image is jointly determined according to the first water body segmentation result and the at least one second water body segmentation result.
  • the first water body segmentation result determined based on the spectral information and the second water body segmentation result determined based on multiple preset water body segmentation methods can be combined, and through the mutual correction between the multiple water body segmentation results,
  • the obtained target water body segmentation results have higher quality, thereby effectively improving the accuracy and accuracy of water body segmentation.
  • step S11 may be flexibly determined according to the actual situation.
  • step S11 may include:
  • Step S111 obtaining the normalized water index of multiple pixels in the first target image according to the spectral information of the first target image;
  • Step S112 taking the pixel points whose values of the normalized water index are within the preset index value range among the plurality of pixel points as the pixel points of the water body area, to obtain the first water body segmentation result of the first target image.
  • the normalized water index may be information obtained by performing normalized interpolation processing on a specific waveband of the first target image. Information on areas of water bodies.
  • the normalized water index of a plurality of pixels in the first target image is obtained, which may be the normalized water index of each pixel in the first target image, or the first target image according to the actual situation.
  • the normalized water index of some pixels in the target image, etc., the specific number of pixels used to obtain the normalized water index can be flexibly selected according to the actual situation, and is not limited in the embodiments of the present disclosure.
  • the method of obtaining the normalized water index of multiple pixels in the first target image is not limited in the embodiments of the present disclosure, and any related method that can calculate NDWI can be used as the implementation form of step S111, and is not limited to the following Various disclosed embodiments.
  • the manner of obtaining the normalized water index can be expressed by the following formula (1):
  • NDWI is the normalized water index of a certain pixel in the first target image
  • Green is the spectral value of the green band corresponding to the pixel
  • NIR is the spectral value of the near-infrared band corresponding to the pixel.
  • step S112 may be used to determine the first water body segmentation result of the first target image.
  • the pixel points whose value of the normalized water index is within the range of the preset index value can be used as the pixel points of the water body area.
  • the specific value of the preset index value range can be flexibly determined according to the actual situation, and is not limited to the following disclosed embodiments.
  • the preset index value range may be greater than zero, that is, in an example, pixels with a value of the normalized water index greater than 0 may be classified as pixels in the water body area, Pixels with a water index value not greater than 0 are classified as background pixels.
  • the water area formed by a plurality of pixels classified as water areas can be used as the first image of the first target image. Water body segmentation results.
  • the obtained first water body segmentation result may include, in addition to the water body area, buildings or Roads and other areas, so in a possible implementation manner, the water body area actually included in the first target image can be regarded as a subset of the first water body segmentation result.
  • FIG. 2 shows a schematic diagram of a water body segmentation method according to an embodiment of the present disclosure. As shown in FIG.
  • a satellite image can be used as the first target image, and NDWI calculation can be performed on the satellite image, and then, according to the NDWI
  • the water body area can be extracted as the foreground from the satellite image, and the foreground mask shown in the upper right corner of Figure 2 can be obtained as the first water body segmentation result.
  • the segmentation result also includes other areas such as buildings or roads.
  • FIG. 3 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • the water body segmentation method may include:
  • Step S111 Acquire normalized water indices of multiple pixels in the first target image according to the spectral information of the first target image.
  • Step S112 taking the pixel points whose values of the normalized water index are within the preset index value range among the plurality of pixel points as the pixel points of the water body area, to obtain the first water body segmentation result of the first target image.
  • step S12 at least one preset water body segmentation method is used to segment the water body region in the first target image to obtain at least one second water body segmentation result.
  • Step S13 Determine a target water body segmentation result of the first target image according to the first water body segmentation result and at least one second water body segmentation result.
  • the normalized water index of a plurality of pixels in the image is obtained, and then according to the numerical relationship between the normalized water index and the preset index value range, determine Pixel points belonging to the water body area, so as to obtain the first water body segmentation result of the first target image.
  • pixel-level water body segmentation can be performed on the water body area in the first target image, and a more comprehensive and clear segmentation boundary can be obtained.
  • the first water body segmentation result so that the target water body segmentation result determined based on the first water body segmentation result has a higher accuracy, and the accuracy of the water body segmentation is improved.
  • step S12 may include:
  • Step S121 according to the water body data in the public map, determine the water body area in the first target image, and obtain the second water body segmentation result. and / or,
  • Step S122 Acquire at least one type of surface coverage data with a resolution within a preset range, determine a water body area in the first target image according to the water body data in the surface coverage data, and obtain at least one second water body segmentation result.
  • Open Map is an open source map, and the map content can be drawn by users according to handheld GPS devices, aerial photography photos, other free content and local knowledge, etc. Therefore, OSM can contain multiple geographic locations around the world
  • the water body area information in the OSM can be used as the water body data in OSM.
  • the data form of the water body data is not limited in the embodiments of the present disclosure.
  • the water body data in the OSM may be: a colored line segment or a shape (polygon or The specific color, line segment and shape, etc., can be selected flexibly according to the actual situation.
  • the water body data in the OSM can include water body area information in multiple geographic locations around the world
  • the water body data corresponding to the geographic range can be searched from the OSM according to the geographic range corresponding to the first target image, and the first target image can be obtained by searching for the water body data corresponding to the geographic range
  • the determined water body area in the first target image is used as the second water body segmentation result.
  • How to determine the geographic range corresponding to the first target image can be flexibly selected according to the actual situation. For details, please refer to the following disclosed embodiments. , which will not be expanded here.
  • the OSM labeling result in the figure is the second water body segmentation result determined according to the water body data in the OSM.
  • the Narrow bodies of water for efficient segmentation based on the second water body segmentation result determined by the OSM.
  • Land cover data can be segmentation results of certain satellite imagery related to the land surface, including water body segmentation results.
  • the quality of the segmentation result of the surface coverage data is related to the resolution of the segmented satellite image.
  • the surface coverage data with a resolution within a preset range can be obtained, and the result of the water body segmentation in the surface coverage data can be obtained.
  • the data of the water body according to the geographical range corresponding to the first target image, the data of the water body corresponding to the geographical range is searched from the surface coverage data to determine the water body area in the first target image, and the determined first target image The water body area in is used as the second water body segmentation result.
  • the low-resolution water body product result in the figure is the second water body segmentation result determined based on the surface coverage data. It can be seen from the figure that the second water body segmentation result determined based on the surface coverage data, It can effectively segment wider water bodies such as rivers, and has a clearer segmentation boundary.
  • the preset range of the resolution can be flexibly set according to the actual situation.
  • a satellite image with a lower resolution can be selected.
  • the surface coverage data in an example, the surface coverage data corresponding to a satellite image with a resolution of 10 meters can be selected.
  • the range covered by the surface coverage data needs to cover the geographic range corresponding to the first target image as much as possible.
  • the data can be global land cover data.
  • At least one type of surface coverage data with a resolution within a preset range can be acquired.
  • a second water body segmentation result can be obtained according to the surface coverage data; if the surface coverage data with multiple resolutions within the preset range is obtained, for each type of surface coverage data, a second water body segmentation result can be obtained;
  • various second water body segmentation results for different land cover data are obtained.
  • the resolutions of different ground cover data may be the same or different, which are not limited in the embodiments of the present disclosure.
  • step S12 may include step S121 and/or step S122, that is, in the process of obtaining the second water body segmentation result, only the second water body segmentation determined based on the water body data in the OSM may be obtained.
  • step S121 may include step S121 and/or step S122, that is, in the process of obtaining the second water body segmentation result, only the second water body segmentation determined based on the water body data in the OSM may be obtained.
  • FIG. 4 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • the water body segmentation method may include:
  • Step S11 segment the water body region in the first target image according to the spectral information of the first target image to obtain a first water body segmentation result of the first target image.
  • Step S121 according to the water body data in the public map, determine the water body area in the first target image, and obtain the second water body segmentation result; and/or,
  • Step S122 Acquire at least one type of surface coverage data with a resolution within a preset range, determine a water body area in the first target image according to the water body data in the surface coverage data, and obtain at least one second water body segmentation result.
  • Step S13 Determine a target water body segmentation result of the first target image according to the first water body segmentation result and at least one second water body segmentation result.
  • multiple second water body segmentation results can be obtained in various ways, and the data between the multiple second water body segmentation results can be complementary to each other to maximize the The water body in the first target image is segmented more comprehensively, thereby improving the overall segmentation accuracy of the obtained multiple second water body segmentation results, and then improving the accuracy of the target water body segmentation results determined based on the multiple second water body segmentation results.
  • step S121 may include:
  • the water body area corresponding to the geographical range in the public map is taken as the target water body area
  • the target water body area is a water body dividing line
  • extend the water body dividing line to a preset width in a preset direction, and use the expanded target water body area as the second water body dividing result
  • the target water body area is a polygonal area and/or a circular area
  • the target water body area is used as the second water body segmentation result.
  • the geographic range corresponding to the first target image may be the geographic area where the water body area in the first target image is located, and how to determine the geographic range corresponding to the first target image can be flexibly determined according to the actual situation of the first target image.
  • the first target image when the first target image is a remote sensing image, the first target image itself may contain geographic information, so the geographic information contained in the first target image may be used to determine the corresponding image of the first target image. geographic range.
  • the water body area corresponding to the geographic range in the public map can be used as the target water body area.
  • the water body data in the OSM can be: covered on the map Colored line segments or shapes (polygons or circles) on the inner water body area, so the target water body area determined from OSM may be a water body dividing line in the form of a line segment, or it may be a non-line segment area.
  • the shape is not limited in the embodiments of the present disclosure, and may be a polygon or a circle, or the like.
  • the form of the target water body area may not be considered, and the target water body area may be directly used as the second water body segmentation result.
  • the obtained second water body segmentation result may be used as the water body region labeling of the first target image to be used for other subsequent water body segmentation processes, such as training a neural network with a water body segmentation function.
  • the second water body segmentation result in the form of a water body segmentation line may not be used as an annotation for neural network training and other processes. Therefore, in a possible implementation manner, the second water body segmentation result can also be obtained according to the form of the target water body area.
  • the form of the target water body area is a polygon and/or a circle, etc. In the case of line segments, the target water body area can be directly used as the second water body segmentation result.
  • the water body dividing line when the target water body area is in the form of a water body dividing line, the water body dividing line can be extended to a predetermined width in a preset direction, so as to obtain an expanded target water body in the form of non-line segments area, the expanded target water body area can be used as the second water body segmentation result.
  • both the preset direction and the preset width can be flexibly set according to the actual situation, which are not limited in the embodiments of the present disclosure.
  • the preset direction may be a certain direction or multiple directions. In the case where the preset direction includes multiple directions, the preset widths in each direction may be the same or different.
  • the water body may be divided into Each point on the line is expanded outward with r as the radius, so as to obtain the polygonal target water body area as the second water body segmentation result.
  • Fig. 5 shows a schematic diagram of expanding a water body dividing line according to an embodiment of the present disclosure. It can be seen from Fig. 5 that the water body dividing line in the form of a line segment in the figure can be expanded into a polygonal area with a certain width through expansion. The polygonal area can be used as the second water body segmentation result.
  • FIG. 6 shows a schematic diagram of annotating the first target image with the expanded second water body segmentation result according to an embodiment of the present disclosure.
  • the second water body segmentation result can also be used as The water body area labeling of the first target image is superimposed with the first target image.
  • the expanded target water body area is As the second water body segmentation result, when the target water body area is a polygonal area and/or a circular area, the target water body area is taken as the second water body segmentation result.
  • step S13 the first water body segmentation result and at least one second water body segmentation result can be obtained, and then the target water body segmentation result of the first target image can be determined through step S13.
  • the implementation manner of step S13 may be flexibly determined according to the actual situation.
  • step S13 may include:
  • Step S131 combining at least one second water body segmentation result to obtain a combined water body segmentation result
  • Step S132 taking the water body region included in the first water body segmentation result and the combined water body segmentation result as the target water body segmentation result of the first target image.
  • the combined water body segmentation result may be a result obtained by superimposing at least one second water body segmentation result.
  • the second water body segmentation result may be determined according to OSM or at least one type of land cover data
  • the segmentation result of the water body based on the OSM can effectively segment the narrower water body, and the segmentation result determined based on the surface cover data can effectively segment the wider water body. Therefore, for different second water bodies
  • the superposition of the segmentation results can play a complementary role, and obtain the combined water body segmentation results with higher accuracy.
  • the second water body segmentation result determined according to the OSM may include an expansion process for the water body segmentation line, and the expansion process may reduce the accuracy of the boundary of the second water body segmentation result obtained after expansion.
  • the second water body segmentation result determined according to the ground cover data may be limited by the accuracy of the ground cover data itself, resulting in a decrease in the boundary precision of the determined second water body segmentation result.
  • the first water body segmentation result has a clear segmentation boundary and includes a relatively comprehensive water body area.
  • the intersection of the first water body segmentation result and the combined water body segmentation result can be obtained to obtain the water body area commonly included in the first water body segmentation result and the combined water body segmentation result, and then, The jointly included water body area can be used as the target water body segmentation result of the first target image, so that the boundary accuracy of the target water body segmentation result can be effectively improved, and then the accuracy and effect of the water body segmentation can be improved.
  • FIG. 7 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure.
  • the water body segmentation method may include:
  • Step S11 segment the water body region in the first target image according to the spectral information of the first target image to obtain a first water body segmentation result of the first target image.
  • step S12 at least one preset water body segmentation method is used to segment the water body region in the first target image to obtain at least one second water body segmentation result.
  • Step S131 combining at least one second water body segmentation result to obtain a combined water body segmentation result
  • Step S132 taking the water body region included in the first water body segmentation result and the combined water body segmentation result as the target water body segmentation result of the first target image.
  • the combined water body segmentation result is obtained by taking the union of at least one second water body segmentation result
  • the target water body segmentation result of the first target image is obtained by taking the intersection of the combined water body segmentation result and the first water body segmentation result.
  • the target water body segmentation result of the first target image has higher accuracy and improves the quality of water body segmentation.
  • the second water body segmentation result can also be used as the water body region labeling of the first target image, and correspondingly, the first target image obtained based on the second water body segmentation result can also be marked.
  • the target water body segmentation result is used to label the first target image. Therefore, FIG. 8 shows a flowchart of a water body segmentation method according to an embodiment of the present disclosure. As shown in FIG. 8 , in a possible implementation manner, the water body segmentation method may further include:
  • Step S11 segment the water body region in the first target image according to the spectral information of the first target image to obtain a first water body segmentation result of the first target image.
  • step S12 at least one preset water body segmentation method is used to segment the water body region in the first target image to obtain at least one second water body segmentation result.
  • Step S13 Determine a target water body segmentation result of the first target image according to the first water body segmentation result and at least one second water body segmentation result.
  • Step S14 according to the target water body segmentation result of the first target image, mark the water body region in the first target image, and obtain the first target image including the label.
  • Step S15 using the marked first target image as a sample to train an initial neural network model to obtain a water body segmentation network.
  • the target water body segmentation result of the first target image obtained by the methods of the above disclosed embodiments may be due to the expansion of the water body segmentation line, resulting in an inaccurate segmentation boundary, or due to the ground surface
  • the accuracy of the coverage data is low, which reduces the accuracy of the segmentation result. Therefore, in order to further improve the accuracy of the water body segmentation, the first target image can also be labeled according to the target water body segmentation result of the first target image, so as to use the labeled image.
  • the first target image is used as a sample to train the initial neural network model to obtain a water body segmentation network that can be used for water body segmentation.
  • the water body region in the first target image is marked, and each water body in the target water body segmentation result of the first target image may be marked in the first target image, It is also possible to select some of the target water body segmentation results for marking. The specific selection can be flexibly determined according to the actual situation, which is not limited in the embodiment of the present disclosure.
  • the obtained first target image including the annotation can be used as a training sample to train an initial neural network model, where the initial neural network model can be any model, and is not limited to the following disclosed embodiments.
  • the initial neural network model may be a semantic segmentation network model, such as U-Net, FC-Densenet, or HRNet.
  • HRNet can be chosen as the initial neural network model.
  • the trained water body segmentation network can be used to perform water body segmentation on the images input into the network. Specifically, how to use the water body segmentation network for application is not limited in the embodiments of the present disclosure. For details, please refer to the following disclosed embodiments. Do not expand first.
  • FIG. 9 shows a flowchart of a method for dividing a water body according to an embodiment of the present disclosure.
  • the method can be applied to a second water body dividing device.
  • the second water body dividing device may be the same device as the first water body dividing device, or may be different devices.
  • the second water body dividing device may be a terminal device, a server, or other processing devices.
  • the terminal device reference may be made to the above disclosed embodiments, which will not be repeated here.
  • the water body segmentation method can be applied to a cloud server or a local server, and the cloud server can be a public cloud server or a private cloud server, which can be flexibly selected according to the actual situation.
  • the water body segmentation method can also be implemented by the processor calling computer-readable instructions stored in the memory.
  • an embodiment of the present disclosure further proposes a water body segmentation method, including:
  • Step S21 the second target image is input into the water body segmentation network to obtain a third water body segmentation result of the second target image.
  • Step S22 segment the water body region of the second target image according to the spectral information of the second target image to obtain a fourth water body segmentation result of the second target image.
  • Step S23 taking the water body region included in the third water body segmentation result and the fourth water body segmentation result as the target water body segmentation result of the second target image.
  • the second target image may be any image that has the requirement of water body segmentation, and its implementation form may refer to the first target image in the above disclosed embodiments, which will not be repeated here.
  • the "first" and “second" in the first target image and the second target image are only used to distinguish which water body segmentation process the image is used in, and do not limit the Whether the implementation manner is the same, in a possible implementation manner, the first target image and the second target image may be the same image, or may be different images.
  • the water body segmentation network may be any neural network with a water body segmentation function, and its implementation form is not limited in the embodiments of the present disclosure.
  • the water body segmentation network may be the one mentioned in the above disclosed embodiment,
  • the specific form of the neural network obtained by training the target water body segmentation result of the first target image can be found in the above disclosed embodiments, which will not be repeated here.
  • the third water body segmentation result may be a result obtained by using a water body segmentation network to perform water body segmentation on the second target image, and its implementation form may also refer to the first water body segmentation result or the second water body segmentation result, which will not be repeated here.
  • step S22 the water body region of the second target image is segmented according to the spectral information of the second target image.
  • the segmentation process reference may be made to the spectral information of the first target image in the above-mentioned disclosed embodiments.
  • the fourth water body dividing result obtained in step S22 reference may also be made to the first water body dividing result or the second water body dividing result in the above disclosed embodiments, which will not be repeated here.
  • step S21 and step S22 are not limited in the embodiment of the present disclosure, that is, the order of performing water body segmentation on the second target image in different ways is not limited in the embodiment of the present disclosure.
  • step S21 and step S22 may be implemented simultaneously, and in a possible implementation manner, steps S21 and S22 and the like may also be implemented respectively according to a preset implementation sequence.
  • the water body region jointly included in the third water body segmentation result and the fourth water body segmentation result may be used as the target water body segmentation result of the second target image.
  • the fourth water body segmentation result can be obtained by segmenting according to the spectral information of the second target image, referring to the acquisition process of the first water body segmentation result, it can be known that the water body area actually included in the second target image can be regarded as the fourth water body area. Subset of water body segmentation results.
  • the water body area commonly included between the third water body segmentation result obtained by the water body segmentation network and the fourth water body segmentation result is used as the target water body segmentation result of the second target image, and the fourth water body segmentation result does not belong to the water body.
  • the segmentation results of the regions are excluded to obtain more accurate segmentation results.
  • the water body segmentation is performed on the second target image through the water body segmentation network and the spectral information of the second target image, respectively, to obtain a third water body segmentation result and a fourth water body segmentation result, so that the third water body segmentation result is divided into The water body region included in the fourth water body segmentation result is used as the target water body segmentation result of the second target image.
  • the method of determining the third water body segmentation result through the water body segmentation network can reduce the problem of confusion between water bodies and buildings, roads, shadows, etc., when the second target image belongs to different data sources or contains large Under the circumstance of the range of the region, it can have a stable water body segmentation effect, so that the obtained water body segmentation results have high accuracy, and the segmentation is more convenient and easy to achieve; on the other hand, using the The four water body segmentation results are intersected with the third water body segmentation results to obtain the target water body segmentation results of the second target image, which can further optimize the boundaries of the obtained water body segmentation results and reduce the false scenes that are incorrectly segmented in the water body segmentation results.
  • the water body segmentation network may be trained by the target water body segmentation result of the first target image, that is, in a possible implementation manner, the water body segmentation network may be based on the first target image.
  • a target image and the target water body segmentation result of the first target image are obtained by training.
  • how to train the water body segmentation network according to the first target image and the target water body segmentation result of the first target image is not limited in the embodiments of the present disclosure.
  • the first target image is marked with the target water body segmentation result of the first target image, and the marked image is used as a sample to input
  • the water body segmentation network is obtained by training the initial neural network model.
  • a third water body segmentation result is obtained by performing water body segmentation on the second target image through a water body segmentation network trained according to the first target image and the target water body segmentation result of the first target image. Since the third water body segmentation result has high segmentation accuracy, it can be seen that through the above process, the water body segmentation network obtained by training can have a better water body segmentation effect, which can not only improve the accuracy of the third water body segmentation result, further, The accuracy of the target water body segmentation result of the second target image can also be improved.
  • the target water body segmentation result of the first target image can be obtained flexibly in different ways. Therefore, in a possible implementation manner, the water body segmentation method proposed by the embodiments of the present disclosure may further include:
  • the water body region in the first target image is segmented to obtain a first water body segmentation result of the first target image
  • the target water body segmentation result of the first target image is determined.
  • the process of obtaining the target water body segmentation result of the first target image used for training the water body segmentation network can be jointly determined based on the spectral information and at least one preset water body segmentation method. Therefore, the training of the water body segmentation network Data can be automatically generated, reducing the cost of manual labeling, while having high accuracy and quality.
  • the loss function can be flexibly selected according to the actual situation.
  • the water body segmentation network is trained through a preset loss function, wherein, in the case of training the water body segmentation network through a preset loss function, the difference between the first training result and the second training result The difference is within the preset difference range, the first training result includes the training result obtained by training with the target water body segmentation result of the first target image as the labeled sample, and the second training result includes the manually input water body segmentation result as the labeled sample.
  • the training results obtained by training samples are provided by training samples.
  • the water body segmentation network can be trained by using a preset loss function.
  • the preset loss function may be a loss function with noise robustness, wherein the loss function with noise robustness may be obtained and not obtained when there is noise in the annotations in the sample (such as inaccurate annotations, etc.).
  • the training results of samples containing noise are similar.
  • the segmentation result of the target water body of the first target image may be due to the expansion of the water body segmentation line or the accuracy of the surface coverage data, etc., resulting in a less clear segmentation boundary.
  • the target water body segmentation result as the labeled sample may contain a certain amount of noise.
  • the labeled samples that is, those labeled with artificial water body segmentation results, can generally be considered to contain no noise.
  • the preset loss function is based on the noise-containing samples (for example, the above-mentioned first target image containing annotations, that is, the first target image using the target water segmentation result of the first target image as the annotated first target image).
  • the difference between the two is the preset Within the range of difference, that is, the two can have a relatively close training effect.
  • the preset difference range can be flexibly determined according to the actual situation, which is not limited in the embodiments of the present disclosure.
  • What kind of loss function with robustness to noise is specifically selected as the preset loss function is not limited in this embodiment of the present disclosure.
  • a GCE loss function and/or an RCE loss function may be selected. , as the default loss function.
  • Using a preset loss function with noise robustness to train the water body segmentation network can further reduce inaccurate or erroneous segmentation results in the target water body segmentation results of the first target image, and learn and train the water body segmentation network. Therefore, the obtained water body segmentation network has higher water body segmentation accuracy, thereby improving the accuracy of the target water body segmentation result of the second target image obtained subsequently.
  • FIG. 10 shows a block diagram of a first water body dividing apparatus according to an embodiment of the present disclosure.
  • the first water body segmentation device 30 may include: a first water body segmentation module 31, configured to segment the water body region in the first target image according to the spectral information of the first target image to obtain the first target The first water body segmentation result of the image;
  • the second water body segmentation module 32 is configured to segment the water body region in the first target image by at least one preset water body segmentation method to obtain at least one second water body segmentation result; the target water body segmentation result determination module 33 is used for According to the first water body segmentation result and the at least one second water body segmentation result, the target water body segmentation result of the first target image is determined.
  • the target water body segmentation result determination module is configured to: combine at least one second water body segmentation result to obtain a combined water body segmentation result;
  • the water body area is used as the target water body segmentation result of the first target image.
  • the second water body segmentation module is configured to: determine the water body region in the first target image according to the water body data in the public map, and obtain the second water body segmentation result; and/or obtain at least one For the surface coverage data with a resolution within a preset range, the water body region in the first target image is determined according to the water body data in the surface coverage data, and at least one second water body segmentation result is obtained.
  • the second water body segmentation module is further configured to: according to the geographical range corresponding to the first target image, take the water body area corresponding to the geographical range in the public map as the target water body area; in the target water body area, In the case of a water body dividing line, extend the water body dividing line to a preset width in a preset direction, and use the expanded target water body area as the second water body segmentation result; if the target water body area is a polygonal area and/or a circular area In this case, the target water body area is used as the second water body segmentation result.
  • the first water body segmentation module is configured to: obtain the normalized water index of multiple pixels in the first target image according to the spectral information of the first target image; The pixel points whose value of the water index is within the range of the preset index value are regarded as the pixel points of the water body area, and the first water body segmentation result of the first target image is obtained.
  • the device is further configured to: mark the water body region in the first target image according to the target water body segmentation result of the first target image, so as to obtain a first target image including the label;
  • the first target image is taken as a sample, and the initial neural network model is trained to obtain a water body segmentation network.
  • FIG. 11 shows a block diagram of a second water body dividing apparatus according to an embodiment of the present disclosure.
  • the second water body segmentation device 40 may include: a third water body segmentation module 41 for inputting the second target image into a water body segmentation network to obtain a third water body segmentation result of the second target image; a fourth water body The segmentation module 42 is used for segmenting the water body region of the second target image according to the spectral information of the second target image to obtain a fourth water body segmentation result of the second target image; the target water body segmentation result acquisition module 43 is used for The water body region included in the three water body segmentation results and the fourth water body segmentation result is used as the target water body segmentation result of the second target image.
  • the water body segmentation network is obtained by training according to the first target image and the target water body segmentation result of the first target image.
  • the device is further configured to: segment the water body region in the first target image according to the spectral information of the first target image to obtain a first water body segmentation result of the first target image; A preset water body segmentation method, in which the water body area in the first target image is segmented to obtain at least one second water body segmentation result; the first target image is determined according to the first water body segmentation result and the at least one second water body segmentation result The target water body segmentation result.
  • the water body segmentation network is trained through a preset loss function, wherein, in the case of training the water body segmentation network through a preset loss function, the difference between the first training result and the second training result The difference is within the preset difference range, the first training result includes the training result obtained by training with the target water body segmentation result of the first target image as the labeled sample, and the second training result includes the manually input water body segmentation result as the labeled sample.
  • the application example of the present disclosure proposes a water body segmentation method, which can segment the water body region in an image with high precision.
  • the water body segmentation method proposed in the application example of the present disclosure may include:
  • the first step is to automate the construction of water body labeling datasets.
  • the NDWI of each pixel in the image can be calculated by formula (1) in the above disclosed embodiment according to the spectrum of the satellite image, and then the pixel whose value is greater than zero is set as For category 1 (that is, the foreground of the water body area), set the pixels whose value is less than or equal to 0 to category 0 (that is, the background), and obtain a foreground mask with a good segmentation boundary as the first water body segmentation result, as shown in the upper right corner of Figure 2.
  • the obtained foreground mask basically includes all the water body areas, but there are also other categories, such as buildings and roads, so the water body can be regarded as a subset of the foreground mask.
  • the second water body segmentation result can also be obtained according to the labeling result of the water body area in the OSM. Since the labeling of the water body area in the OSM is divided into two categories, one is the polygon labeling, as shown in the shaded part in Figure 5. , the polygon label can be directly obtained and used for subsequent water body labeling of satellite images. The other type is water body segmentation line labeling, as shown in the line segment in Figure 5. This water body segmentation line cannot be directly used as the water body labeling for subsequent satellite images.
  • the water body dividing line can be expanded and expanded outward with a radius of r to obtain a polygonal result, and then the polygonal annotation and the expanded polygonal result of the water body dividing line can be superimposed to obtain an OSM-based
  • the second water body segmentation result that is, the OSM labeling result in Figure 2).
  • the problem of rough boundary may occur.
  • the second water body segmentation result can also be obtained from the low-resolution water body product.
  • a 10-meter resolution global surface coverage product (10 types in total) can be selected as the surface coverage data, and the surface coverage data can be divided into The data of the middle water body is extracted separately as the second water body segmentation result (ie, the low-resolution water body product result in Figure 2).
  • the second water body segmentation result can better extract the wider river, and the boundary is also Finer, as shown in the display box of the low-resolution water product results in Figure 2, but narrower rivers (as shown in the display box of the OSM annotation results in Figure 2) cannot be extracted.
  • the second water body segmentation result based on OSM can be superimposed with the second water body segmentation result obtained based on the low-resolution water body product to obtain the combined water body segmentation result, so that the narrower water body labeling in OSM and the low-resolution water body can be combined.
  • the wider river annotation in the rate water product plays a complementary role and can greatly improve the quality of the annotation dataset.
  • intersection of the first water body segmentation result and the combined water body segmentation result is taken, on the one hand, the intersection can be used as the water body segmentation result of the satellite image (that is, the target water body segmentation result of the first target image mentioned in the above disclosed embodiments), and on the other hand, the intersection can be used as the water body segmentation result of the satellite image.
  • it can be used as the water body area labeling of satellite images (that is, the data set constructed in Figure 2) to construct an automated water body labeling data set, thereby effectively improving the boundary effect of water body area labeling, as shown in the display box in the lower right corner of Figure 2. Show.
  • the expansion and expansion of the water body dividing line and the inaccuracy of the low-resolution water body product result itself may lead to a decrease in the accuracy of the intersection result.
  • the second step is automated water body segmentation.
  • Water body segmentation can be regarded as a two-class problem (two types of background and water).
  • the method of deep learning semantic segmentation can be used, and the image in the automatic water body labeling dataset obtained in the first step can be used as input, A water body segmentation network is obtained by performing supervised learning with labeled data as labels.
  • any common semantic segmentation network can be used, such as U-Net, FC-Densenet, etc.
  • the semantic segmentation network HRNet with better effect can be selected.
  • a loss function with noise robust properties can also be used to train the water body segmentation network, as described in the present disclosure.
  • any loss function with noise robustness can be used, such as GCE loss function or RCE loss function. The impact of water body segmentation network learning.
  • the image to be subjected to water body segmentation (that is, the second target image in the above disclosed embodiments) can be input into the water body segmentation network to obtain a third water body segmentation result.
  • the third water body segmentation result output by the water body segmentation network can be post-processed.
  • NDWI can be used to extract the foreground mask of the second target image for water body segmentation to obtain the fourth water body.
  • the segmentation result is obtained, and then the intersection of the third water body segmentation result and the fourth water body segmentation result is obtained to obtain the target water body segmentation result of the second target image, so that the boundary of the water body segmentation result can be further optimized, and the virtual scene in the water body segmentation result can be suppressed.
  • the water body segmentation method proposed in the application example of the present disclosure can not only be applied to the segmentation of water body regions, but also can be further extended to be applied to the segmentation of other object regions, such as soil or buildings.
  • the sampling segmentation method can be flexibly changed with different objects, and is not limited to the above disclosed embodiments.
  • the writing order of each step does not mean a strict execution order but constitutes any limitation on the implementation process, and the specific execution order of each step should be based on its function and possible Internal logic is determined.
  • Embodiments of the present disclosure further provide a computer-readable storage medium, on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the foregoing method is implemented.
  • the computer-readable storage medium may be a volatile computer-readable storage medium or a non-volatile computer-readable storage medium.
  • An embodiment of the present disclosure further provides an electronic device, including: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to perform the above method.
  • the above-mentioned memory can be a volatile memory (volatile memory), such as RAM; or a non-volatile memory (non-volatile memory), such as ROM, flash memory (flash memory), hard disk (Hard Disk Drive) , HDD) or solid-state drive (Solid-State Drive, SSD); or a combination of the above types of memory, and provide instructions and data to the processor.
  • volatile memory such as RAM
  • non-volatile memory such as ROM, flash memory (flash memory), hard disk (Hard Disk Drive) , HDD) or solid-state drive (Solid-State Drive, SSD); or a combination of the above types of memory, and provide instructions and data to the processor.
  • the above-mentioned processor can be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, controller, microcontroller, and microprocessor. It can be understood that, for different devices, the electronic device used to implement the function of the processor may also be other, which is not specifically limited in the embodiment of the present disclosure.
  • the electronic device may be provided as a terminal, server or other form of device.
  • an embodiment of the present disclosure further provides a computer program, which implements the above method when the computer program is executed by a processor.
  • FIG. 12 is a block diagram of an electronic device 800 according to an embodiment of the present disclosure.
  • electronic device 800 may be a mobile phone, computer, digital broadcast terminal, messaging device, game console, tablet device, medical device, fitness device, personal digital assistant, etc. terminal.
  • an electronic device 800 may include one or more of the following components: a processing component 802, a memory 804, a power supply component 806, a multimedia component 808, an audio component 810, an input/output (I/O) interface 812, a sensor component 814 , and the communication component 816 .
  • the processing component 802 generally controls the overall operation of the electronic device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 802 can include one or more processors 820 to execute instructions to perform all or some of the steps of the methods described above.
  • processing component 802 may include one or more modules that facilitate interaction between processing component 802 and other components.
  • processing component 802 may include a multimedia module to facilitate interaction between multimedia component 808 and processing component 802.
  • Memory 804 is configured to store various types of data to support operation at electronic device 800 . Examples of such data include instructions for any application or method operating on electronic device 800, contact data, phonebook data, messages, pictures, videos, and the like. Memory 804 may be implemented by any type of volatile or nonvolatile storage device or combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable Programmable Read Only Memory (EPROM), Programmable Read Only Memory (PROM), Read Only Memory (ROM), Magnetic Memory, Flash Memory, Magnetic or Optical Disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read only memory
  • EPROM erasable Programmable Read Only Memory
  • PROM Programmable Read Only Memory
  • ROM Read Only Memory
  • Magnetic Memory Flash Memory
  • Magnetic or Optical Disk Magnetic Disk
  • Power supply assembly 806 provides power to various components of electronic device 800 .
  • Power supply components 806 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to electronic device 800 .
  • Multimedia component 808 includes a screen that provides an output interface between the electronic device 800 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from a user.
  • the touch panel includes one or more touch sensors to sense touch, swipe, and gestures on the touch panel. The touch sensor may not only sense the boundaries of a touch or swipe action, but also detect the duration and pressure associated with the touch or swipe action.
  • the multimedia component 808 includes a front-facing camera and/or a rear-facing camera. When the electronic device 800 is in an operation mode, such as a shooting mode or a video mode, the front camera and/or the rear camera may receive external multimedia data. Each of the front and rear cameras can be a fixed optical lens system or have focal length and optical zoom capability.
  • Audio component 810 is configured to output and/or input audio signals.
  • audio component 810 includes a microphone (MIC) that is configured to receive external audio signals when electronic device 800 is in operating modes, such as calling mode, recording mode, and voice recognition mode.
  • the received audio signal may be further stored in memory 804 or transmitted via communication component 816 .
  • audio component 810 also includes a speaker for outputting audio signals.
  • the I/O interface 812 provides an interface between the processing component 802 and a peripheral interface module, which may be a keyboard, a click wheel, a button, or the like. These buttons may include, but are not limited to: home button, volume buttons, start button, and lock button.
  • Sensor assembly 814 includes one or more sensors for providing status assessment of various aspects of electronic device 800 .
  • the sensor assembly 814 can detect the on/off state of the electronic device 800, the relative positioning of the components, such as the display and the keypad of the electronic device 800, the sensor assembly 814 can also detect the electronic device 800 or one of the electronic device 800 Changes in the position of components, presence or absence of user contact with the electronic device 800 , orientation or acceleration/deceleration of the electronic device 800 and changes in the temperature of the electronic device 800 .
  • Sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects in the absence of any physical contact.
  • Sensor assembly 814 may also include a light sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor assembly 814 may also include an acceleration sensor, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • Communication component 816 is configured to facilitate wired or wireless communication between electronic device 800 and other devices.
  • Electronic device 800 may access wireless networks based on communication standards, such as WiFi, 2G or 3G, or a combination thereof.
  • the communication component 816 receives broadcast signals or broadcast related personnel information from an external broadcast management system via a broadcast channel.
  • the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication.
  • NFC near field communication
  • the NFC module may be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • electronic device 800 may 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 A programmed gate array (FPGA), controller, microcontroller, microprocessor or other electronic component implementation is used to perform the above method.
  • ASICs application specific integrated circuits
  • DSPs digital signal processors
  • DSPDs digital signal processing devices
  • PLDs programmable logic devices
  • FPGA field programmable A programmed gate array
  • controller microcontroller, microprocessor or other electronic component implementation is used to perform the above method.
  • a non-volatile computer-readable storage medium such as a memory 804 comprising computer program instructions executable by the processor 820 of the electronic device 800 to perform the above method is also provided.
  • FIG. 13 is a block diagram of an electronic device 1900 according to an embodiment of the present disclosure.
  • the electronic device 1900 may be provided as a server.
  • electronic device 1900 includes a processing component 1922, which further includes one or more processors, and a memory resource, represented by memory 1932, for storing instructions executable by processing component 1922, such as applications.
  • An application program stored in memory 1932 may include one or more modules, each corresponding to a set of instructions.
  • the processing component 1922 is configured to execute instructions to perform the above-described methods.
  • the electronic device 1900 may also include a power supply assembly 1926 configured to perform power management of the electronic device 1900, a wired or wireless network interface 1950 configured to connect the electronic device 1900 to a network, and an input output (I/O) interface 1958 .
  • Electronic device 1900 may operate based on an operating system stored in memory 1932, such as Windows ServerTM, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM or the like.
  • a non-volatile computer-readable storage medium such as memory 1932 comprising computer program instructions executable by processing component 1922 of electronic device 1900 to perform the above-described method.
  • the present disclosure may be a system, method and/or computer program product.
  • the computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the present disclosure.
  • a computer-readable storage medium may be a tangible device that can hold and store instructions for use by the instruction execution device.
  • the computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing.
  • Non-exhaustive list of computer readable storage media include: portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM) or flash memory), static random access memory (SRAM), portable compact disk read only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, mechanically coded devices, such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.
  • RAM random access memory
  • ROM read only memory
  • EPROM erasable programmable read only memory
  • flash memory static random access memory
  • SRAM static random access memory
  • CD-ROM compact disk read only memory
  • DVD digital versatile disk
  • memory sticks floppy disks
  • mechanically coded devices such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.
  • Computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (eg, light pulses through fiber optic cables), or through electrical wires transmitted electrical signals.
  • the computer readable program instructions described herein may be downloaded to various computing/processing devices from a computer readable storage medium, or to an external computer or external storage device over a network such as the Internet, a local area network, a wide area network, and/or a wireless network.
  • the network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers.
  • a network adapter card or network interface in each computing/processing device receives computer-readable program instructions from a network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .
  • Computer program instructions for carrying out operations of the present disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or instructions in one or more programming languages.
  • Source or object code written in any combination, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as the "C" language or similar programming languages.
  • the computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement.
  • the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (eg, using an Internet service provider through the Internet connect).
  • electronic circuits such as programmable logic circuits, field programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), are personalized by utilizing state personnel information of computer readable program instructions.
  • Computer readable program instructions can be executed to implement various aspects of the present disclosure.
  • These computer readable program instructions may be provided to a processor of a general purpose computer, special purpose computer or other programmable data processing apparatus to produce a machine that causes the instructions when executed by the processor of the computer or other programmable data processing apparatus , resulting in means for implementing the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • These computer readable program instructions can also be stored in a computer readable storage medium, these instructions cause a computer, programmable data processing apparatus and/or other equipment to operate in a specific manner, so that the computer readable medium on which the instructions are stored includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks of the flowchart and/or block diagrams.
  • Computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other equipment to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other equipment to produce a computer-implemented process , thereby causing instructions executing on a computer, other programmable data processing apparatus, or other device to implement the functions/acts specified in one or more blocks of the flowcharts and/or block diagrams.
  • each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more functions for implementing the specified logical function(s) executable instructions.
  • the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved.
  • each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations can be implemented in dedicated hardware-based systems that perform the specified functions or actions , or can be implemented in a combination of dedicated hardware and computer instructions.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Analysis (AREA)

Abstract

本公开涉及一种水体分割方法及装置、电子设备和存储介质。所述方法包括：根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；根据所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。

Description

水体分割方法及装置、电子设备和存储介质
相关公开的交叉引用
本专利申请要求于2020年10月13日提交的、申请号为202011092192.5、发明名称为“水体分割方法及装置、电子设备和存储介质”的中国专利申请的优先权，该申请的全文以引用的方式并入本文中。
技术领域
本公开涉及计算机视觉领域，尤其涉及一种水体分割方法及装置、电子设备和存储介质。
背景技术
水体分割可以从遥感影像中识别并分割出水体区域，是遥感影像处理中的一个重点问题。然而水体分割中往往要面临很多问题，比如容易误检，经常容易与建筑物、道路及阴影等混淆，或是很难提取较窄的水体等。因此，如何获得具有较高质量的水体分割结果，成为目前一个亟待解决的问题。
发明内容
本公开提出了一种水体分割方案。
根据本公开的一方面，提供了一种水体分割方法，包括：根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。
在一种可能的实现方式中，所述根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果，包括：对所述至少一个第二水体分割结果进行合并，得到合并水体分割结果；将所述第一目标图像的所述第一水体分割结果和所述合并水体分割结果中共同包含的水体区域，作为所述第一目标图像的目标水体分割结果。
在一种可能的实现方式中，所述通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果，包括以下中的至少一个：根据公开地图中的水体数据，确定所述第一目标图像中的水体区域，得到第二水体分割结果；获取至少一种分辨率在预设范围内的地表覆盖数据，根据所述地表覆盖数据中水体的数据，确定所述第一目标图像中的水体区域，得到至少一种第二水体分割结果。
在一种可能的实现方式中，所述根据公开地图中的水体数据，确定所述第一目标图像中的水体区域，得到第二水体分割结果，包括：根据所述第一目标图像对应的地理范围，将所述公开地图中与所述地理范围对应的水体区域，作为目标水体区域；在所述目标水体区域为水体分割线的情况下，在预设方向上将所述水体分割线扩展至预设宽度，将扩展后的目标水体区域作为所述第二水体分割结果；在所述目标水体区域为多边形区域和/或圆形区域的情况下，将所述目标水体区域作为所述第二水体分割结果。
在一种可能的实现方式中，所述根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果，包括：根据所述第一目标图像的光谱信息，获取所述第一目标图像中多个像素点的归一化水指数；将所述多个像素点中所述归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点，得到所述第一目标图像的第一水体分割结果。
在一种可能的实现方式中，所述方法还包括：根据所述第一目标图像的目标水体分割结果，对所述第一目标图像中的水体区域进行标注，得到包含标注的第一目标图像；将所述包含标注的第一目标图像作为样本，对初始神经网络模型进行训练，得到水体分割网络。
根据本公开的一方面，提供了一种水体分割方法，包括：将第二目标图像输入水体分割网络，得到所述第二目标图像的第三水体分割结果；根据所述第二目标图像的光谱信息，对所述第二目标图像的水体区域进行分割，得到所述第二目标图像的第四水体分割结果；将所述第二目标图像的所述第三水体分割结果和所述第四水体分割结果中共同包含的水体区域，作为所述第二目标图像的目标水体分割结果。
在一种可能的实现方式中，所述水体分割网络为根据第一目标图像以及所述第一目标图像的目标水体分割结果进行训练得到的。
在一种可能的实现方式中，所述方法还包括：根据所述第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。
在一种可能的实现方式中，所述根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果，包括：对所述至少一个第二水体分割结果进行合并，得到合并水体分割结果；将所述第一目标图像的所述第一水体分割结果和所述合并水体分割结果中共同包含的水体区域，作为所述第一目标图像的目标水体分割结果。
在一种可能的实现方式中，所述通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果，包括以下中的至少一个：根据公开地图中的水体数据，确定所述第一目标图像中的水体区域，得到第二水体分割结果；获取至少一种分辨率在预设范围内的地表覆盖数据，根据所述地表覆盖数据中水体的数据，确定所述第一目标图像中的水体区域，得到至少一种第二水体分割结果。
在一种可能的实现方式中，所述根据公开地图中的水体数据，确定所述第一目标图像中的水体区域，得到第二水体分割结果，包括：根据所述第一目标图像对应的地理范围，将所述公开地图中与所述地理范围对应的水体区域，作为目标水体区域；在所述目标水体区域为水体分割线的情况下，在预设方向上将所述水体分割线扩展至预设宽度，将扩展后的目标水体区域作为所述第二水体分割结果；在所述目标水体区域为多边形区域和/或圆形区域的情况下，将所述目标水体区域作为所述第二水体分割结果。
在一种可能的实现方式中，所述根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果，包括：根据所述第一目标图像的光谱信息，获取所述第一目标图像中多个像素点的归一化水指数；将所述多个像素点中所述归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点，得到所述第一目标图像的第一水体分割结果。
在一种可能的实现方式中，所述方法还包括：根据所述第一目标图像的目标水体分割结果，对所述第一目标图像中的水体区域进行标注，得到包含标注的第一目标图像；将所述包含标注的第一目标图像作为样本，对初始神经网络模型进行训练，得到水体分割网络。
在一种可能的实现方式中，所述水体分割网络通过预设损失函数进行训练，其中，在通过所述预设损失函数对所述水体分割网络进行训练的情况下，第一训练结果和第二训练结果之间的差异在预设差异范围以内，所述第一训练结果包括以所述第一目标图像的目标水体分割结果作为标注的样本进行训练所得到的训练结果，所述第二训练结果包括以人工输入的水体分割结果作为标注的样本进行训练所得到的训练结果。
根据本公开的一方面，提供了一种水体分割装置，包括：第一水体分割模块，用于根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；第二水体分割模块，用于通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；目标水体分割结果确定模块，用于根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。
根据本公开的一方面，提供了一种水体分割装置，包括：第三水体分割模块，用于将第二目标图像输入水体分割网络，得到所述第二目标图像的第三水体分割结果；第四水体分割模块，用于根据所述第二目标图像的光谱信息，对所述第二目标图像的水体区域进行分割，得到所述第二目标图像的第四水体分割结果；目标水体分割结果获取模块，用于将所述第二目标图像的所述第三水体分割结果和所述第四水体分割结果中共同包含的水体区域，作为所述第二目标图像的目标水体分割结果。
根据本公开的一方面，提供了一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为：执行上述水体分割方法。
根据本公开的一方面，提供了一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述水体分割方法。
在本公开实施例中，通过根据光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果，以及通过至少一种预设水体分割方式对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果，从而根据第一水体分割结果和至少一个第二水体分割结果来共同确定第一目标图像的目标水体分割结果。通过上述过程，可以将基于光谱信息所确定的第一水体分割结果，和基于多种预设水体分割方式所确定的第二水体分割结果进行结合，通过多种水体分割结果之间的相互修正，使得得到的目标水体分割结果具有更高的质量，从而有效提升水体分割的精度和准确性。
应当理解的是，以上的一般描述和后文的细节描述仅是示例性和解释性的，而非限制本公开。根据下面参考附图对示例性实施例的详细说明，本公开的其它特征及方面将变得清楚。
附图说明
此处的附图被并入说明书中并构成本说明书的一部分，这些附图示出了符合本公开的实施例，并与说明书一起用于说明本公开的技术方案。
图1示出根据本公开一实施例的水体分割方法的流程图。
图2示出根据本公开一实施例的水体分割方法的示意图。
图3示出根据本公开一实施例的水体分割方法的流程图。
图4示出根据本公开一实施例的水体分割方法的流程图。
图5示出根据本公开一实施例的对水体分割线进行扩展的示意图。
图6示出根据本公开一实施例的利用扩展后的第二水体分割结果对第一目标图像进行标注的示意图。
图7示出根据本公开一实施例的水体分割方法的流程图。
图8示出根据本公开一实施例的水体分割方法的流程图。
图9示出根据本公开一实施例的水体分割方法的流程图。
图10示出根据本公开一实施例的第一水体分割装置的框图。
图11示出根据本公开一实施例的第二水体分割装置的框图。
图12示出根据本公开实施例的一种电子设备的框图。
图13示出根据本公开实施例的一种电子设备的框图。
具体实施方式
以下将参考附图详细说明本公开的各种示例性实施例、特征和方面。附图中相同的附图标记表示功能相同或相似的元件。尽管在附图中示出了实施例的各种方面，但是除非特别指出，不必按比例绘制附图。
在这里专用的词“示例性”意为“用作例子、实施例或说明性”。这里作为“示例性”所说明的任 何实施例不必解释为优于或好于其它实施例。
本文中术语“和/或”，仅仅是一种描述关联对象的关联关系，表示可以存在三种关系，例如，A和/或B，可以表示：单独存在A，同时存在A和B，单独存在B这三种情况。另外，本文中术语“至少一种”表示多种中的任意一种或多种中的至少两种的任意组合，例如，包括A、B、C中的至少一种，可以表示包括从A、B和C构成的集合中选择的任意一个或多个元素。
另外，为了更好地说明本公开，在下文的具体实施方式中给出了众多的具体细节。本领域技术人员应当理解，没有某些具体细节，本公开同样可以实施。在一些实例中，对于本领域技术人员熟知的方法、手段、元件和电路未作详细描述，以便于凸显本公开的主旨。
图1示出根据本公开一实施例的水体分割方法的流程图，该方法可以应用于第一水体分割装置，第一水体分割装置可以为终端设备、服务器或者其他处理设备等。其中，终端设备可以为用户设备(User Equipment，UE)、移动设备、用户终端、终端、蜂窝电话、无绳电话、个人数字处理(Personal Digital Assistant，PDA)、手持设备、计算设备、车载设备、可穿戴设备等。在一个示例中，该水体分割方法可以应用于云端服务器或本地服务器，云端服务器可以为公有云服务器，也可以为私有云服务器，根据实际情况灵活选择即可。
在一些可能的实现方式中，该水体分割方法也可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。
如图1所示，在一种可能的实现方式中，所述水体分割方法可以包括：
步骤S11，根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果。
步骤S12，通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果。
步骤S13，根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
其中，第一目标图像可以是任一具有水体分割需求的图像，其实现形式在本公开实施例中不做限制。在一种可能的实现方式中，第一目标图像可以是遥感影像，遥感影像的分辨率可以根据获取的影像的实际情况灵活决定，在一个示例中，第一目标图像可以为具有较高分辨率的4波段遥感影像。
第一目标图像的来源以及获取方式在本公开实施例中也不做限制，在一种可能的实现方式中，可以获取相关硬件设备直接采集的遥感影像，在一种可能的实现方式中，也可以从存储有遥感影像的数据库中，读取或选定其中的部分或全部影像，来作为第一目标图像。
第一目标图像的光谱信息，可以是第一目标图像中本身包含的多个波段的光谱信息，第一目标图像中包含哪些波段的光谱信息，可以根据第一目标图像的实际情况灵活决定，在本公开实施例中不做限制。在一种可能的实现方式中，第一目标图像中可以包含可见光波段(RGB)以及近红外波段(NIR，Near Infrared)的光谱信息；在一种可能的实现方式中，第一目标图像在包含RGB和NIR波段的光谱信息的基础上，还可以包含更多其他波段的光谱信息。
步骤S11中，由于图像中不同的对象的光谱信息可能存在差异，因此可以根据第一目标图像的光谱信息，来对第一目标图像中的水体区域进行分割。步骤S11的具体实现方式可以根据实际情况灵活决定，详见下述各公开实施例，在此先不做展开。第一水体分割结果可以包含第一目标图像中水体区域的位置以及形状等信息，分割后得到的第一水体分割结果可以通过直观的图像方式进行展示，也可以通过信息的方式进行存储，其实现方式可以根据实际情况灵活决定，在本公开实施例中不做限制。
步骤S12中，通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割的实现方式同样可以根据实际情况灵活决定，同样可以参见下述各公开实施例，在此先不做展开。其中，预设水体分割方式可以是任意的可以对图像中的水体区域进行分割的方式，比如通过相关的水体分割计算方法，或是通过查找包含水体分割结果的某些数据库等，亦或是通过具有水 体分割功能的神经网络等，具体采用哪种或哪些方式可以根据实际情况灵活决定。在一种可能的实现方式中，具体通过几种预设水体分割方式来对第一目标图像进行水体分割，其数量在本公开实施例中同样不做限制。随着预设水体分割方式的不同，得到的第二水体分割结果也可能有差别，因此在一种可能的实现方式中，第二水体分割结果的数量可以与预设水体分割方式的种类对应，在本公开实施例中不做限制。第二水体分割结果的实现形式可以参考第一水体分割结果，在此不再赘述，第一水体分割结果与第二水体分割结果的实现形式可以相同，也可以不同，在本公开实施例中不做限制。需要注意的是，在本公开实施例中，第一水体分割结果与第二水体分割结果中的“第一”和“第二”等仅用于区分通过不同方式所得到的水体分割结果，并不限定水体分割结果的获取顺序等，后续实施例中水体分割结果的其他编号同理，不再赘述。
步骤S11与步骤S12的实现顺序在本公开实施例中不做限制，即通过不同方式对第一目标图像进行水体分割的顺序在本公开实施例中不做限定，在一种可能的实现方式中，步骤S11与步骤S12可以同时实现，在一种可能的实现方式中，也可以按照预设的实现顺序，分别实现步骤S11与S12等。
在得到第一水体分割结果和至少一个第二水体分割结果后，可以根据第一水体分割结果以及至少一个第二水体分割，来确定第一目标图像的目标水体分割结果。其中，目标水体分割结果可以是目标图像中水体区域的最终确定的分割结果，由于第一水体分割结果和第二水体分割结果均可能存在水体分割不准确的情况，因此在一种可能的实现方式中，基于第一水体分割结果和至少一个第二水体分割结果之间的差异，在第一水体分割结果和/或第二水体分割结果的基础上进行调整，可以得到更为准确的目标水体分割结果。目标水体分割结果的实现形式可以参考上述公开实施例中的第一水体分割结果以及第二水体分割结果等，在此不再赘述。如何基于第一水体分割结果以及至少一个第二水体分割结果，来确定目标水体分割结果，可以详见下述各公开实施例，在此先不做展开。
在本公开实施例中，根据光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果，以及通过至少一种预设水体分割方式对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果，从而根据第一水体分割结果和至少一个第二水体分割结果来共同确定第一目标图像的目标水体分割结果。通过上述过程，可以将基于光谱信息所确定的第一水体分割结果，和基于多种预设水体分割方式所确定的第二水体分割结果进行结合，通过多种水体分割结果之间的相互修正，使得得到的目标水体分割结果具有更高的质量，从而有效提升水体分割的精度和准确性。
如上述各公开实施例所述，步骤S11的实现方式可以根据实际情况灵活决定。在一种可能的实现方式中，步骤S11可以包括：
步骤S111，根据第一目标图像的光谱信息，获取第一目标图像中多个像素点的归一化水指数；
步骤S112，将多个像素点中归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点，得到第一目标图像的第一水体分割结果。
其中，归一化水指数(NDWI，Normalized Difference Water Index)可以是对第一目标图像的特定波段进行归一化插值处理所得到的信息，根据归一化水指数，可以凸显第一目标图像中水体区域的信息。
步骤S111中，获取第一目标图像中多个像素点的归一化水指数，可以是获取第一目标图像中每个像素点的归一化水指数，也可以是根据实际情况，获取第一目标图像中部分像素点的归一化水指数等，用于获取归一化水指数的像素点的具体数量，可以根据实际情况灵活选择，在本公开实施例中不做限制。
获取第一目标图像中多个像素点的归一化水指数的方式在本公开实施例中不做限制，任何可以计算NDWI的相关方法，均可以作为步骤S111的实现形式，不局限于下述各公开实施例。在一种可能的实现方式中，获取归一化水指数的方式可以通过下述公式(1)表示：
NDWI＝(Green-NIR)/(Green+NIR)      (1)
其中，NDWI为第一目标图像中某个像素点的归一化水指数，Green为该像素点对应的绿色波段的光谱值，NIR为该像素点对应的近红外波段的光谱值。
获取多个像素点的归一化水指数后，可以通过步骤S112，来确定第一目标图像的第一水体分割结果。如上述公开实施例所述，可以将归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点。其中，预设指数值范围的具体数值可以根据实际情况灵活决定，不局限于下述各公开实施例。在一种可能的实现方式中，预设指数值范围可以为大于零，即在一个示例中，可以将归一化水指数的值大于0的像素点归类为水体区域的像素点，将归一化水指数的值不大于0的像素点归类为背景像素点，在这种情况下，多个归类为水体区域的像素点所构成的水体区域，可以作为第一目标图像的第一水体分割结果。
由于第一目标图像中可能存在“异物同谱”的情况，即不同类型的地物呈现出相同的光谱特征，因此得到的第一水体分割结果中除了包含水体区域以外，还可能包括建筑物或道路等其他区域，因此在一种可能的实现方式中，第一目标图像中实际包含的水体区域，可以看作为第一水体分割结果的子集。图2示出根据本公开一实施例的水体分割方法的示意图，如图2所示，在一个示例中，可以将卫星影像作为第一目标图像，对卫星影像进行NDWI计算，然后，根据NDWI的计算结果，可以从卫星影像中将水体区域作为前景进行提取，则可以得到如图2右上角所示的前景掩模作为第一水体分割结果，从图2中可以看出，得到的第一水体分割结果中除了包含水体区域以外，也包含建筑或道路等其他区域。
基于上述各公开实施例，图3示出根据本公开一实施例的水体分割方法的流程图，如图3所示，在一种可能的实现方式中，水体分割方法可以包括：
步骤S111，根据第一目标图像的光谱信息，获取第一目标图像中多个像素点的归一化水指数。
步骤S112，将多个像素点中归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点，得到第一目标图像的第一水体分割结果。
步骤S12，通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果。
步骤S13，根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
在本公开实施例中，根据第一目标图像的光谱信息，获取图像中多个像素点的归一化水指数，继而根据归一化水指数与预设指数值范围之间的数值关系，确定属于水体区域的像素点，从而得到第一目标图像的第一水体分割结果，通过上述过程，可以对第一目标图像中的水体区域进行像素级别的水体分割，得到具有清晰分割边界且较为全面的第一水体分割结果，从而使得基于第一水体分割结果确定的目标水体分割结果具有更高的准确度，提升水体分割的精度。
在一种可能的实现方式中，步骤S12可以包括：
步骤S121，根据公开地图中的水体数据，确定第一目标图像中的水体区域，得到第二水体分割结果。和/或，
步骤S122，获取至少一种分辨率在预设范围内的地表覆盖数据，根据地表覆盖数据中水体的数据，确定第一目标图像中的水体区域，得到至少一种第二水体分割结果。
其中，公开地图(OSM，OpenStreetMap)为开源地图，其中的地图内容可以由用户根据手持GPS设备、航空摄影照片、其他自由内容以及本地知识等进行绘制，因此，OSM中可以包含全球多个地理位置中的水体区域信息，这些水体区域信息可以作为OSM中的水体数据。水体数据的数据形式在本公开实施例中不做限制，在一种可能的实现方式中，OSM中的水体数据可以为：覆盖于地图内水体区域上的具有颜色的线段或是形状(多边形或圆形)等，具体采用何种颜色以及何种线段与形状等，均可以根据实际情况灵活选择。
由于OSM中的水体数据可以包含全球多个地理位置的水体区域信息，因此，可以根据第一目标图像对应的地理范围，来从OSM中查找对应该地理范围的水体数据，得到第一目标图像中 的水体区域，并将确定出的第一目标图像中的水体区域作为第二水体分割结果，具体如何确定第一目标图像对应的地理范围可以根据实际情况灵活选择，详见下述各公开实施例，在此先不做展开。如图2所示，图中的OSM标注结果即为根据OSM中的水体数据所确定的第二水体分割结果，从图2中可以看出，基于OSM所确定的第二水体分割结果，可以对较窄的水体进行有效的分割。
地表覆盖数据可以是与地表相关的某些卫星影像的分割结果，其中包含水体分割结果。地表覆盖数据的分割结果的质量与其分割的卫星影像的分辨率相关，在一种可能的实现方式中，可以获取分辨率在预设范围内的地表覆盖数据，将地表覆盖数据中水体分割的结果作为水体的数据，从而根据第一目标图像对应的地理范围，从地表覆盖数据中查找对应地理范围的水体的数据，来确定第一目标图像中的水体区域，并将确定出的第一目标图像中的水体区域作为第二水体分割结果。如图2所示，图中的低分辨率水体产品结果即为根据地表覆盖数据所确定的第二水体分割结果，从图中可以看出，基于地表覆盖数据所确定的第二水体分割结果，可以对较宽的水体如河流等进行有效的分割，且具有较为清晰的分割边界。
分辨率的预设范围可以根据实际情况灵活设定，在一种可能的实现方式中，为了与基于OSM所得到的第二水体分割结果进行互补，可以选用分辨率较低的卫星影像所对应的地表覆盖数据，在一个示例中，可以选用10米分辨率的卫星影像所对应的地表覆盖数据。
为了在地表覆盖数据中查找到第一目标图像对应地理范围的水体的数据，地表覆盖数据覆盖的范围需要尽可能覆盖第一目标图像对应的地理范围，在一种可能的实现方式中，地表覆盖数据可以为全球地表覆盖数据。
通过上述公开实施例可以看出，在一种可能的实现方式中，可以获取至少一种分辨率在预设范围内的地表覆盖数据，在这种情况下，如果仅获取一种分辨率在预设范围内的地表覆盖数据，则可以根据该地表覆盖数据来得到一种第二水体分割结果；如果获取多种分辨率在预设范围内的地表覆盖数据，则针对每种地表覆盖数据，可以分别通过上述公开实施例中提到的得到第二水体分割结果的方法，得到多种针对于不同的地表覆盖数据的第二水体分割结果。在获取多种地表覆盖数据的情况下，不同地表覆盖数据的分辨率可以相同，也可以不同，在本公开实施例中不做限制。
通过上述公开实施例还可以看出，步骤S12可以包括步骤S121和/或步骤S122，即在得到第二水体分割结果的过程中，可以仅得到基于OSM中的水体数据所确定的第二水体分割结果，也可以仅得到基于地表覆盖数据所确定的一个或多个第二水体分割结果，还可以同时得到基于OSM中的水体数据所确定的第二水体分割结果，以及基于地表数据所确定的一个或多个第二水体分割结果等，具体如何选择，可以根据实际情况灵活选择。
基于上述各公开实施例，图4示出根据本公开一实施例的水体分割方法的流程图，如图4所示，在一种可能的实现方式中，水体分割方法可以包括：
步骤S11，根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果。
步骤S121，根据公开地图中的水体数据，确定第一目标图像中的水体区域，得到第二水体分割结果；和/或，
步骤S122，获取至少一种分辨率在预设范围内的地表覆盖数据，根据地表覆盖数据中水体的数据，确定第一目标图像中的水体区域，得到至少一种第二水体分割结果。
步骤S13，根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
通过上述公开实施例中提出的步骤S12的实现过程，可以采用多种不同的方式，来获取多个第二水体分割结果，多个第二水体分割结果之间的数据可以相互补充，来尽可能地对第一目标图像中的水体进行较为全面地分割，从而提升得到的多个第二水体分割结果的总体分割精度，继而提升基于多个第二水体分割结果所确定的目标水体分割结果的精度。
在一种可能的实现方式中，步骤S121可以包括：
根据第一目标图像对应的地理范围，将公开地图中与该地理范围对应的水体区域，作为目标水体区域；
在目标水体区域为水体分割线的情况下，在预设方向上将水体分割线扩展至预设宽度，将扩展后的目标水体区域作为第二水体分割结果；
在目标水体区域为多边形区域和/或圆形区域的情况下，将目标水体区域作为第二水体分割结果。
其中，第一目标图像对应的地理范围，可以是第一目标图像中水体区域所在的地理区域，具体如何确定第一目标图像对应的地理范围，可以根据第一目标图像的实际情况灵活决定。在一种可能的实现方式中，在第一目标图像为遥感影像的情况下，第一目标图像本身可以包含地理信息，因此可以根据第一目标图像包含的地理信息来确定第一目标图像对应的地理范围。
在确定第一目标图像对应的地理范围以后，可以将公开地图中与该地理范围对应的水体区域，作为目标水体区域，上述公开实施例中已经提出，OSM中的水体数据可以为：覆盖于地图内水体区域上的具有颜色的线段或是形状(多边形或圆形)等，因此从OSM中确定的目标水体区域，可能为线段形式的水体分割线，也可能为非线段的区域，该区域的形状在本公开实施例中不做限制，可以为多边形或是圆形等。
在一种可能的实现方式中，可以不考虑目标水体区域的形式，直接将目标水体区域作为第二水体分割结果。
在一种可能的实现方式中，得到的第二水体分割结果可能会作为第一目标图像的水体区域标注，来用于后续其他的水体分割过程，比如训练具有水体分割功能的神经网络等。在这种情况下，形式为水体分割线的第二水体分割结果，可能无法作为标注来用于神经网络的训练等过程。因此，在一种可能的实现方式中，还可以根据目标水体区域的形式来得到第二水体分割结果，如上述公开实施例所述，在目标水体区域的形式为多边形和/或圆形等非线段的情况下，可以直接将目标水体区域作为第二水体分割结果。在一种可能的实现方式中，在目标水体区域的形式为水体分割线的情况下，可以在预设方向上将水体分割线扩展至预设宽度，从而得到扩展后的非线段形式的目标水体区域，该扩展后的目标水体区域可以作为第二水体分割结果。其中，预设方向和预设宽度均可以根据实际情况灵活设定，在本公开实施例中不做限制。预设方向可以为某一方向，也可以为多个方向，在预设方向包括多个方向的情况下，各方向上的预设宽度可以相同也可以不同，在一个示例中，可以对水体分割线上的各点均进行以r为半径的向外扩展，从而得到多边形的目标水体区域，作为第二水体分割结果。图5示出根据本公开一实施例的对水体分割线进行扩展的示意图，从图5中可以看出，图中线段形式的水体分割线可以通过扩展膨胀为一具有一定宽度的多边形区域，该多边形区域可以作为第二水体分割结果。其中，受限于周边环境和实际地貌等，多边形区域沿所述水体分割线的长度方向上的宽度有变化，但多边形区域的整体延伸姿势是与水体分割线大体一致的。图6示出根据本公开一实施例的利用扩展后的第二水体分割结果对第一目标图像进行标注的示意图，如图6所示，在一个示例中，该第二水体分割结果还可以作为第一目标图像的水体区域标注，与第一目标图像进行叠加。
通过将公开地图中与第一目标图像对应的地理范围相匹配的水体区域作为目标水体区域，并在目标水体区域为水体分割线的情况下对水体分割线进行扩展，将扩展后的目标水体区域作为第二水体分割结果，在目标水体区域为多边形区域和/或圆形区域的情况下将目标水体区域作为第二水体分割结果，通过上述过程，可以根据公开地图中目标水体区域的不同形式，灵活地得到相应的第二水体分割结果，从而与后续利用第二水体分割结果所执行的更多水体分割操作相适应。
通过上述各公开实施例，可以得到第一水体分割结果和至少一个第二水体分割结果，继而可以通过步骤S13来确定第一目标图像的目标水体分割结果。步骤S13的实现方式可以根据实际情况灵活决定，在一种可能的实现方式中，步骤S13可以包括：
步骤S131，对至少一个第二水体分割结果进行合并，得到合并水体分割结果；
步骤S132，将第一水体分割结果和合并水体分割结果中共同包含的水体区域，作为第一目 标图像的目标水体分割结果。
其中，合并水体分割结果可以是对至少一个第二水体分割结果进行叠加所得到的结果，如上述公开实施例所述，第二水体分割结果可以是根据OSM或是至少一种地表覆盖数据所确定的水体分割结果，其中，基于OSM确定的分割结果可以对较窄的水体进行有效分割，基于地表覆盖数据所确定的分割结果可以对较宽的水体进行有效分割，因此，对不同的第二水体分割结果进行叠加可以起到互补的作用，得到具有较高精度的合并水体分割结果。
在一种可能的实现方式中，根据OSM确定的第二水体分割结果中可能包含对水体分割线的扩展过程，该扩展过程可能会导致扩展后得到的第二水体分割结果的边界的精度降低。同时，根据地表覆盖数据所确定的第二水体分割结果，可能会受到地表覆盖数据自身的精度限制，而导致确定的第二水体分割结果的边界精度降低。而且，上述公开实施例中还提出，第一水体分割结果具有清晰的分割边界且包含的水体区域较为全面。结合上述内容，在一种可能的实现方式中，可以通过对第一水体分割结果和合并水体分割结果取交集，来得到第一水体分割结果和合并水体分割结果中共同包含的水体区域，然后，可以将共同包含的水体区域作为第一目标图像的目标水体分割结果，从而能够有效提升目标水体分割结果的边界精度，继而提升水体分割的精度和效果。
基于上述各公开实施例，图7示出根据本公开一实施例的水体分割方法的流程图，如图所示，在一种可能的实现方式中，水体分割方法可以包括：
步骤S11，根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果。
步骤S12，通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果。
步骤S131，对至少一个第二水体分割结果进行合并，得到合并水体分割结果；
步骤S132，将第一水体分割结果和合并水体分割结果中共同包含的水体区域，作为第一目标图像的目标水体分割结果。
综上所述，通过对至少一个第二水体分割结果取并集得到合并水体分割结果，对合并水体分割结果和第一水体分割结果取交集得到第一目标图像的目标水体分割结果，可以使得得到的第一目标图像的目标水体分割结果具有更高的精度，提升水体分割的质量。
上述公开实施例中提到过，在一个示例中，第二水体分割结果还可以作为第一目标图像的水体区域标注，相应的，也可以基于第二水体分割结果所得到的第一目标图像的目标水体分割结果，对第一目标图像进行标注。因此，图8示出根据本公开一实施例的水体分割方法的流程图，如图8所示，在一种可能的实现方式中，水体分割方法还可以包括：
步骤S11，根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果。
步骤S12，通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果。
步骤S13，根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
步骤S14，根据第一目标图像的目标水体分割结果，对第一目标图像中的水体区域进行标注，得到包含标注的第一目标图像。
步骤S15，将包含标注的第一目标图像作为样本，对初始神经网络模型进行训练，得到水体分割网络。
在一种可能的实现方式中，通过上述各公开实施例的方法所得到的第一目标图像的目标水体分割结果，可能会因为水体分割线的扩展导致分割的边界不精细，或是由于基于地表覆盖数据的精度较低而降低分割结果的准确度，因此，为进一步提高水体分割的准确度，还可以 根据第一目标图像的目标水体分割结果对第一目标图像进行标注，从而利用标注后的第一目标图像作为样本，来对初始神经网络模型进行训练，得到可以用于进行水体分割的水体分割网络。
其中，根据第一目标图像的目标水体分割结果，对第一目标图像中的水体区域进行标注，可以是将第一目标图像的目标水体分割结果中的各水体均标注在第一目标图像中，也可以是选定其中的部分目标水体分割结果来进行标注，具体如何选择，可以根据实际情况灵活决定，在本公开实施例中不做限制。
得到的包含标注的第一目标图像可以作为训练用样本，对初始神经网络模型进行训练，其中，初始神经网络模型可以是任意模型，不局限于下述各公开实施例。在一种可能的实现方式中，初始神经网络模型可以是语义分割网络模型，比如U-Net、FC-Densenet或是HRNet等。在一个示例中，可以选择HRNet作为初始神经网络模型。
训练得到的水体分割网络可以用于对输入到网络中的图像进行水体分割，具体如何利用水体分割网络进行应用，在本公开实施例中不做限制，详见下述各公开实施例，在此先不做展开。
图9示出根据本公开一实施例的水体分割方法的流程图，该方法可以应用于第二水体分割装置，第二水体分割装置可以与第一水体分割装置为同一装置，也可以为不同装置，第二水体分割装置可以为终端设备、服务器或者其他处理设备等。其中，终端设备的实现形式可以参考上述各公开实施例，在此不再赘述。在一个示例中，该水体分割方法可以应用于云端服务器或本地服务器，云端服务器可以为公有云服务器，也可以为私有云服务器，根据实际情况灵活选择即可。
在一些可能的实现方式中，该水体分割方法也可以通过处理器调用存储器中存储的计算机可读指令的方式来实现。
如图9所示，在一种可能的实现方式中，本公开实施例还提出一种水体分割方法，包括：
步骤S21，将第二目标图像输入水体分割网络，得到第二目标图像的第三水体分割结果。
步骤S22，根据第二目标图像的光谱信息，对第二目标图像的水体区域进行分割，得到第二目标图像的第四水体分割结果。
步骤S23，将第三水体分割结果和第四水体分割结果中共同包含的水体区域，作为第二目标图像的目标水体分割结果。
其中，第二目标图像可以是具有水体分割需求的任意图像，其实现形式可以参考上述各公开实施例中的第一目标图像，在此不再赘述。需要注意的是，第一目标图像与第二目标图像中的“第一”和“第二”等仅用于区分用于该图像应用在何种水体分割过程之中，并不限制该图像的实现方式是否相同，在一种可能的实现方式中，第一目标图像与第二目标图像可以为相同图像，也可以为不同图像。
水体分割网络可以是任意具有水体分割功能的神经网络，其实现形式在本公开实施例中不做限制，在一种可能的实现方式中，水体分割网络可以是上述公开实施例中提到的，通过第一目标图像的目标水体分割结果进行训练所得到的神经网络，其具体形式详见上述各公开实施例，在此不再赘述。
第三水体分割结果可以是利用水体分割网络对第二目标图像进行水体分割所得到的结果，其实现形式也可以参考上述第一水体分割结果或第二水体分割结果，在此不再赘述。
步骤S22中，根据第二目标图像的光谱信息，对第二目标图像的水体区域进行分割，其分割过程可以参考上述公开实施例中根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割的过程，类似的，步骤S22得到的第四水体分割结果同样可以参考上述公开实施例中的第一水体分割结果或第二水体分割结果，在此不再赘述。
步骤S21和步骤S22的执行顺序在本公开实施例中也不做限制，即通过不同方式对第二目标图像进行水体分割的顺序在本公开实施例中不做限定，在一种可能的实现方式中，步骤 S21与步骤S22可以同时实现，在一种可能的实现方式中，也可以按照预设的实现顺序，分别实现步骤S21与S22等。
在得到第三水体分割结果和第四水体分割结果后，可以将第三水体分割结果和第四水体分割结果共同包含的水体区域，作为第二目标图像的目标水体分割结果。其中，由于第四水体分割结果可以根据第二目标图像的光谱信息来进行分割所得到，参考第一水体分割结果的获取过程可知，第二目标图像中实际包含的水体区域，可以看作为第四水体分割结果的子集。因此将通过水体分割网络所得到的第三水体分割结果与第四水体分割结果之间共同包含的水体区域来作为第二目标图像的目标水体分割结果，可以将第四水体分割结果中不属于水体区域的分割结果进行排除，来得到较为准确的分割结果。
在本公开实施例中，分别通过水体分割网络和第二目标图像的光谱信息，对第二目标图像进行水体分割，得到第三水体分割结果和第四水体分割结果，从而将第三水体分割结果和第四水体分割结果中共同包含的水体区域作为第二目标图像的目标水体分割结果。通过上述过程，一方面通过水体分割网络确定第三水体分割结果的方式，可以减少水体与建筑物、道路、阴影等内容的混淆问题，在第二目标图像属于不同的数据源或是包含较大的区域范围的情况下，都能有稳定的水体分割效果，从而使得得到的水体分割结果具有较高的准确度，且分割较为便捷，易于实现；另一方面，利用基于光谱信息所得到的第四水体分割结果，与第三水体分割结果取交集，来得到第二目标图像的目标水体分割结果，可以进一步优化得到的水体分割结果的边界，减少水体分割结果中误分割的虚景等。
如上述公开实施例所述，在一种可能的实现方式中，水体分割网络可以通过第一目标图像的目标水体分割结果训练，即在一种可能的实现方式中，水体分割网络可以为根据第一目标图像以及第一目标图像的目标水体分割结果进行训练得到的。
其中，如何根据第一目标图像和第一目标图像的目标水体分割结果来训练得到水体分割网络，在本公开实施例中不做限制。在一种可能的实现方式中，可以参考上述公开实施例中的步骤S14和S15，利用第一目标图像的目标水体分割结果对第一目标图像进行标注，并将标注后的图像作为样本，输入至初始神经网络模型来训练得到水体分割网络，具体过程详见上述各公开实施例，在此不再赘述。
通过根据第一目标图像以及第一目标图像的目标水体分割结果进行训练得到的水体分割网络，对第二目标图像进行水体分割得到第三水体分割结果。由于第三水体分割结果具有较高的分割精度，因此可知通过上述过程，可以使得训练得到的水体分割网络具有较好的水体分割效果，不仅能够提升第三水体分割结果的准确性，进一步的，还能提升第二目标图像的目标水体分割结果的准确度。
如上述公开实施例所述，第一目标图像的目标水体分割结果可以通过不同的方式所灵活得到，因此在一种可能的实现方式中，本公开实施例提出的水体分割方法还可以包括：
根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果；
通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；
根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
上述过程的具体实现方式可以参考前述各公开实施例，在此不再赘述。通过上述过程，可以看出，用于训练水体分割网络的第一目标图像的目标水体分割结果的得到过程，可以基于光谱信息和至少一种预设水体分割方式共同确定，因此水体分割网络的训练数据可以自动化生成，减少了人工标注的成本，同时具有较高的准确度和质量。
在对水体分割网络进行训练的过程中，可以根据实际情况灵活选择损失函数。在一种可能的实现方式中，水体分割网络通过预设损失函数进行训练，其中，在通过预设损失函数对水体分割网络进行训练的情况下，第一训练结果和第二训练结果之间的差异在预设差异范围以 内，第一训练结果包括以第一目标图像的目标水体分割结果作为标注的样本进行训练所得到的训练结果，第二训练结果包括以人工输入的水体分割结果作为标注的样本进行训练所得到的训练结果。
通过上述公开实施例可以看出，在一种可能的实现方式中，可以通过预设损失函数对水体分割网络进行训练。其中，预设损失函数可以为具有噪声鲁棒性质的损失函数，其中具有噪声鲁棒性质的损失函数可以在样本中的标注存在噪声(比如存在不准确的标注等)的情况下，取得与不包含噪声的样本相近的训练结果。上述公开实施例中已经提到，第一目标图像的目标水体分割结果可能由于水体分割线的扩展或是地表覆盖数据的精度等原因，导致分割的边界不太清晰，因此以第一目标图像的目标水体分割结果作为标注的样本、也即标注有基于本公开方案获得的目标水体分割结果的样本，可能包含一定的噪声，而由于人工进行水体分割较为准确，因此以人工输入的水体分割结果作为标注的样本、也即标注了人工水体分割结果的一般可以认为不包含噪声。如上述公开实施例所述，预设损失函数在基于该包含噪声的样本(例如，上述包含标注的第一目标图像，也即以第一目标图像的目标水体分割结果作为标注的第一目标图像)训练所得到的第一训练结果，与基于不包含噪声的样本(例如，以人工输入的水体分割结果作为标注的图像)训练所得到的第二训练结果，二者之间的差异在预设差异范围以内，即二者可以具有较为接近的训练效果。其中，预设差异范围可以根据实际情况灵活决定，在本公开实施例中不做限定。
具体选用何种具有噪声鲁棒性质的损失函数，来作为预设损失函数，在本公开实施例中不做限制，在一种可能的实现方式中，可以选用GCE损失函数和/或RCE损失函数，来作为预设损失函数。
通过具有噪声鲁棒性质的预设损失函数，来对水体分割网络训练，可以进一步减少第一目标图像的目标水体分割结果中不准确或是有错误的分割结果，对水体分割网络进行学习训练的影响，从而使得得到的水体分割网络具有更高的水体分割精度，继而提升后续得到的第二目标图像的目标水体分割结果的精度。
图10示出根据本公开一实施例的第一水体分割装置的框图。如图所示，所述第一水体分割装置30可以包括：第一水体分割模块31，用于根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果；
第二水体分割模块32，用于通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；目标水体分割结果确定模块33，用于根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
在一种可能的实现方式中，目标水体分割结果确定模块用于：对至少一个第二水体分割结果进行合并，得到合并水体分割结果；将第一水体分割结果和合并水体分割结果中共同包含的水体区域，作为第一目标图像的目标水体分割结果。
在一种可能的实现方式中，第二水体分割模块用于：根据公开地图中的水体数据，确定第一目标图像中的水体区域，得到第二水体分割结果；和/或，获取至少一种分辨率在预设范围内的地表覆盖数据，根据地表覆盖数据中水体的数据，确定第一目标图像中的水体区域，得到至少一种第二水体分割结果。
在一种可能的实现方式中，第二水体分割模块进一步用于：根据第一目标图像对应的地理范围，将公开地图中与地理范围对应的水体区域，作为目标水体区域；在目标水体区域为水体分割线的情况下，在预设方向上将水体分割线扩展至预设宽度，将扩展后的目标水体区域作为第二水体分割结果；在目标水体区域为多边形区域和/或圆形区域的情况下，将目标水体区域作为第二水体分割结果。
在一种可能的实现方式中，第一水体分割模块用于：根据第一目标图像的光谱信息，获取第一目标图像中多个像素点的归一化水指数；将多个像素点中归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点，得到第一目标图像的第一水体分割结果。
在一种可能的实现方式中，该装置还用于：根据第一目标图像的目标水体分割结果， 对第一目标图像中的水体区域进行标注，得到包含标注的第一目标图像；将包含标注的第一目标图像作为样本，对初始神经网络模型进行训练，得到水体分割网络。
图11示出根据本公开一实施例的第二水体分割装置的框图。如图所示，所述第二水体分割装置40可以包括：第三水体分割模块41，用于将第二目标图像输入水体分割网络，得到第二目标图像的第三水体分割结果；第四水体分割模块42，用于根据第二目标图像的光谱信息，对第二目标图像的水体区域进行分割，得到第二目标图像的第四水体分割结果；目标水体分割结果获取模块43，用于将第三水体分割结果和第四水体分割结果中共同包含的水体区域，作为第二目标图像的目标水体分割结果。
在一种可能的实现方式中，水体分割网络为根据第一目标图像以及第一目标图像的目标水体分割结果进行训练得到的。
在一种可能的实现方式中，该装置还用于：根据第一目标图像的光谱信息，对第一目标图像中的水体区域进行分割，得到第一目标图像的第一水体分割结果；通过至少一种预设水体分割方式，对第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；根据第一水体分割结果，以及至少一个第二水体分割结果，确定第一目标图像的目标水体分割结果。
在一种可能的实现方式中，水体分割网络通过预设损失函数进行训练，其中，在通过预设损失函数对水体分割网络进行训练的情况下，第一训练结果和第二训练结果之间的差异在预设差异范围以内，第一训练结果包括以第一目标图像的目标水体分割结果作为标注的样本进行训练所得到的训练结果，第二训练结果包括以人工输入的水体分割结果作为标注的样本进行训练所得到的训练结果。
应用场景示例
对于高分辨率的卫星遥感影像，如何对影像中的水体区域进行准确的分割成为目前一个亟待解决的问题。
本公开应用示例提出了一种水体分割方法，可以对图像中的水体区域进行高精度的分割。
参考上述公开实施例中提出的图2、图5及图6，本公开应用示例中提出的水体分割方法可以包括：
第一步，自动化水体标注数据集构建。
如图2所示，在本公开应用示例中，首先可以根据卫星影像的光谱，通过上述公开实施例中的公式(1)计算影像中各像素点的NDWI，然后将值大于零的像素点设为类别1(即水体区域前景)，将值小于等于0的像素点设为类别0(即背景)，得到有良好的分割边界的前景掩膜作为第一水体分割结果，如图2右上角所示，可以看出，得到的前景掩膜基本包含所有的水体区域，但同时也有其他的类别，如建筑物和道路，因此可以将水体看作为该前景掩膜的子集。
除此之外，还可以根据OSM中水体区域的标注结果来得到第二水体分割结果，由于OSM中水体区域的标注分为两类，一类是多边形标注，如图5中阴影覆盖部分所示，该多边形标注可以直接获取，用于后续对卫星影像的水体标注，另一类是水体分割线标注，如图5中线段部分所示，该水体分割线无法直接作为后续对卫星影像的水体标注，因此在本公开应用示例中，可以对水体分割线进行扩展膨胀，以半径为r向外扩充，得到多边形结果，然后将多边形标注和水体分割线扩展后的多边形结果进行叠加，得到基于OSM的第二水体分割结果(即图2中的OSM标注结果)。在本公开应用示例中，由于对水体分割线进行扩展的过程中，无法确定河流等水体区域的边界的具体位置，因此可能会产生边界粗糙的问题。
另外，还可以从低分辨率水体产品中获取第二水体分割结果，在本公开应用示例中，可以选择10米分辨率全球地表覆盖产品(共有10类)来作为地表覆盖数据，将地表覆盖数据中水体这一类数据单独提取出来，作为第二水体分割结果(即图2中的低分辨率水体产品结果)，该第二水体分割结果可以对较宽的河流进行较好地提取，边界也较精细，如图2中低分辨率水体产品结果的展示框内所示，但较窄的河流(如图2中OSM标注结果的展示框内所示)无法提 取出来。
进一步地，可以将基于OSM的第二水体分割结果与基于低分辨率水体产品得到的第二水体分割结果进行叠加，得到合并水体分割结果，从而可以结合OSM中对于较窄的水体标注和低分辨率水体产品中较宽的河流标注，如图2所示，两者起到互补的作用，可以大大提高标注数据集的质量。
将第一水体分割结果和合并水体分割结果取交集，该交集一方面可以作为该卫星影像的水体分割结果(即上述各公开实施例中提到的第一目标图像的目标水体分割结果)，另一方面可以作为卫星影像的水体区域标注(即图2中构建的数据集)，来构建自动化水体标注数据集，从而有效的改善水体区域标注的边界效果，如图2中右下角的展示框所示。在本公开应用示例中，水体分割线的膨胀扩展，以及低分辨率水体产品结果本身的不精确，均可能导致该交集结果的精度降低。通过以上的处理，可以完成标注数据集的自动构建，尽量减少人工标注的过程。
第二步，自动化的水体分割。
水体分割可以看作是二分类问题(背景和水两类)，在本公开应用示例中，可以使用深度学习语义分割的方法，将第一步中得到的自动化水体标注数据集中的图像作为输入，用标注数据作为标签进行有监督学习，来得到水体分割网络。在本公开应用示例中，可以使用任意常用语义分割网络，如U-Net，FC-Densenet等，在一个示例中，可以选用效果较好的语义分割网络HRNet。此外，除了使用语义分割常用的交叉熵损失函数，考虑到自动化水体标注数据集中的标注数据可能包含错误标注，还可以使用具有噪声鲁棒性质的损失函数来对水体分割网络进行训练，在本公开应用示例中可以使用任意具有噪声鲁棒性的损失函数，如GCE损失函数或RCE损失函数等，在一个示例中，可以选用效果较好的RCE损失函数用于监督，从而可以有效的减少噪声对水体分割网络学习的影响。
在训练得到水体分割网络后，可以将待进行水体分割的图像(即上述各公开实施例中的第二目标图像)输入到水体分割网络中，得到第三水体分割结果，在一个示例中，还可以对水体分割网络输出的第三水体分割结果进行后处理，与自动化水体标注数据集构建的过程相似，可以先利用NDWI提取待进行水体分割的第二目标图像的前景掩膜，得到第四水体分割结果，再对第三水体分割结果和第四水体分割结果取交集，得到第二目标图像的目标水体分割结果，从而可以进一步优化水体分割结果的边界，抑制水体分割结果中的虚景。
本公开应用示例中提出的水体分割方法，除了可以应用于对水体区域的分割，还可以进一步扩展应用于对其他对象区域的分割，比如土壤或建筑等，在用于对其他对象的分割过程中，采样的分割方式可以随着对象的不同灵活发生变化，不局限于上述各公开实施例。
可以理解，本公开提及的上述各个方法实施例，在不违背原理逻辑的情况下，均可以彼此相互结合形成结合后的实施例，限于篇幅，本公开不再赘述。
本领域技术人员可以理解，在具体实施方式的上述方法中，各步骤的撰写顺序并不意味着严格的执行顺序而对实施过程构成任何限定，各步骤的具体执行顺序应当以其功能和可能的内在逻辑确定。
本公开实施例还提出一种计算机可读存储介质，其上存储有计算机程序指令，所述计算机程序指令被处理器执行时实现上述方法。计算机可读存储介质可以是易失性计算机可读存储介质或非易失性计算机可读存储介质。
本公开实施例还提出一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器被配置为上述方法。
在实际应用中，上述存储器可以是易失性存储器(volatile memory)，例如RAM；或者非易失性存储器(non-volatile memory)，例如ROM，快闪存储器(flash memory)，硬盘(Hard Disk Drive，HDD)或固态硬盘(Solid-State Drive，SSD)；或者上述种类的存储器的组合，并向处理器提供指令和数据。
上述处理器可以为ASIC、DSP、DSPD、PLD、FPGA、CPU、控制器、微控制器、微 处理器中的至少一种。可以理解地，对于不同的设备，用于实现上述处理器功能的电子器件还可以为其它，本公开实施例不作具体限定。
电子设备可以被提供为终端、服务器或其它形态的设备。
基于前述实施例相同的技术构思，本公开实施例还提供了一种计算机程序，该计算机程序被处理器执行时实现上述方法。
图12是根据本公开实施例的一种电子设备800的框图。例如，电子设备800可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等终端。
参照图12，电子设备800可以包括以下一个或多个组件：处理组件802，存储器804，电源组件806，多媒体组件808，音频组件810，输入/输出(I/O)的接口812，传感器组件814，以及通信组件816。
处理组件802通常控制电子设备800的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件802可以包括一个或多个处理器820来执行指令，以完成上述的方法的全部或部分步骤。此外，处理组件802可以包括一个或多个模块，便于处理组件802和其他组件之间的交互。例如，处理组件802可以包括多媒体模块，以方便多媒体组件808和处理组件802之间的交互。
存储器804被配置为存储各种类型的数据以支持在电子设备800的操作。这些数据的示例包括用于在电子设备800上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器804可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电源组件806为电子设备800的各种组件提供电力。电源组件806可以包括电源管理系统，一个或多个电源，及其他与为电子设备800生成、管理和分配电力相关联的组件。
多媒体组件808包括在所述电子设备800和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板(TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件808包括一个前置摄像头和/或后置摄像头。当电子设备800处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件810被配置为输出和/或输入音频信号。例如，音频组件810包括一个麦克风(MIC)，当电子设备800处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器804或经由通信组件816发送。在一些实施例中，音频组件810还包括一个扬声器，用于输出音频信号。
I/O接口812为处理组件802和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件814包括一个或多个传感器，用于为电子设备800提供各个方面的状态评估。例如，传感器组件814可以检测到电子设备800的打开/关闭状态，组件的相对定位，例如所述组件为电子设备800的显示器和小键盘，传感器组件814还可以检测电子设备800或电子设备800一个组件的位置改变，用户与电子设备800接触的存在或不存在，电子设备800方位或加速/减速和电子设备800的温度变化。传感器组件814可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件814还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件814还可以包括 加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件816被配置为便于电子设备800和其他设备之间有线或无线方式的通信。电子设备800可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信组件816经由广播信道接收来自外部广播管理系统的广播信号或广播相关人员信息。在一个示例性实施例中，所述通信组件816还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，电子设备800可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述方法。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器804，上述计算机程序指令可由电子设备800的处理器820执行以完成上述方法。
图13是根据本公开实施例的一种电子设备1900的框图。例如，电子设备1900可以被提供为一服务器。参照图13，电子设备1900包括处理组件1922，其进一步包括一个或多个处理器，以及由存储器1932所代表的存储器资源，用于存储可由处理组件1922的执行的指令，例如应用程序。存储器1932中存储的应用程序可以包括一个或一个以上的每一个对应于一组指令的模块。此外，处理组件1922被配置为执行指令，以执行上述方法。
电子设备1900还可以包括一个电源组件1926被配置为执行电子设备1900的电源管理，一个有线或无线网络接口1950被配置为将电子设备1900连接到网络，和一个输入输出(I/O)接口1958。电子设备1900可以操作基于存储在存储器1932的操作系统，例如Windows ServerTM，Mac OS XTM，UnixTM,LinuxTM，FreeBSDTM或类似。
在示例性实施例中，还提供了一种非易失性计算机可读存储介质，例如包括计算机程序指令的存储器1932，上述计算机程序指令可由电子设备1900的处理组件1922执行以完成上述方法。
本公开可以是系统、方法和/或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于使处理器实现本公开的各个方面的计算机可读程序指令。
计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是――但不限于――电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。
这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算/处理设备，或者通过网络、例如因特网、局域网、广域网和/或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和/或边缘服务器。每个计算/处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算/处理设备中的计算机可读存储介质中。
用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言—诸如Smalltalk、C++等，以及常规的过程式编程语言—诸如“C”语言或类似的编程语言。计算机可读程序指令可以完 全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络—包括局域网(LAN)或广域网(WAN)—连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态人员信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。
这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和/或框图描述了本公开的各个方面。应当理解，流程图和/或框图的每个方框以及流程图和/或框图中各方框的组合，都可以由计算机可读程序指令实现。
这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理器，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理器执行时，产生了实现流程图和/或框图中的一个或多个方框中规定的功能/动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和/或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和/或框图中的一个或多个方框中规定的功能/动作的各个方面的指令。
也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和/或框图中的一个或多个方框中规定的功能/动作。
附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。

Claims (14)

  1. 一种水体分割方法，包括：
    根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；
    通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；
    根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。
  2. 根据权利要求1所述的方法，其特征在于，所述根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果，包括：
    对所述至少一个第二水体分割结果进行合并，得到合并水体分割结果；
    将所述第一目标图像的所述第一水体分割结果和所述合并水体分割结果中共同包含的水体区域，作为所述第一目标图像的目标水体分割结果。
  3. 根据权利要求1或2所述的方法，其特征在于，所述通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果，包括以下中的至少一个：
    根据公开地图中的水体数据，确定所述第一目标图像中的水体区域，得到第二水体分割结果；
    获取至少一种分辨率在预设范围内的地表覆盖数据，根据所述地表覆盖数据中水体的数据，确定所述第一目标图像中的水体区域，得到至少一种第二水体分割结果。
  4. 根据权利要求3所述的方法，其特征在于，所述根据公开地图中的水体数据，确定所述第一目标图像中的水体区域，得到第二水体分割结果，包括：
    根据所述第一目标图像对应的地理范围，将所述公开地图中与所述地理范围对应的水体区域，作为目标水体区域；
    在所述目标水体区域为水体分割线的情况下，在预设方向上将所述水体分割线扩展至预设宽度，将扩展后的目标水体区域作为所述第二水体分割结果；
    在所述目标水体区域为多边形区域和/或圆形区域的情况下，将所述目标水体区域作为所述第二水体分割结果。
  5. 根据权利要求1至4中任意一项所述的方法，其特征在于，所述根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果，包括：
    根据所述第一目标图像的光谱信息，获取所述第一目标图像中多个像素点的归一化水指数；
    将所述多个像素点中所述归一化水指数的值在预设指数值范围内的像素点作为水体区域的像素点，得到所述第一目标图像的第一水体分割结果。
  6. 根据权利要求1至5中任意一项所述的方法，其特征在于，所述方法还包括：
    根据所述第一目标图像的目标水体分割结果，对所述第一目标图像中的水体区域进行标注，得到包含标注的第一目标图像；
    将所述包含标注的第一目标图像作为样本，对初始神经网络模型进行训练，得到水体分割网络。
  7. 一种水体分割方法，包括：
    将第二目标图像输入水体分割网络，得到所述第二目标图像的第三水体分割结果；
    根据所述第二目标图像的光谱信息，对所述第二目标图像的水体区域进行分割，得到所述第二目标图像的第四水体分割结果；
    将所述第二目标图像的所述第三水体分割结果和所述第四水体分割结果中共同包含的水体 区域，作为所述第二目标图像的目标水体分割结果。
  8. 根据权利要求7所述的方法，其特征在于，所述水体分割网络为根据第一目标图像以及所述第一目标图像的目标水体分割结果进行训练得到的。
  9. 根据权利要求8所述的方法，其特征在于，所述方法还包括：
    根据所述第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；
    通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；
    根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。
  10. 根据权利要求8至9任一所述的方法，其特征在于，在通过预设损失函数对所述水体分割网络进行训练的情况下，第一训练结果和第二训练结果之间的差异在预设差异范围以内，所述第一训练结果包括以所述第一目标图像的目标水体分割结果作为标注的样本进行训练所得到的训练结果，所述第二训练结果包括以人工输入的水体分割结果作为标注的样本进行训练所得到的训练结果。
  11. 一种水体分割装置，包括：
    第一水体分割模块，用于根据第一目标图像的光谱信息，对所述第一目标图像中的水体区域进行分割，得到所述第一目标图像的第一水体分割结果；
    第二水体分割模块，用于通过至少一种预设水体分割方式，对所述第一目标图像中的水体区域进行分割，得到至少一个第二水体分割结果；
    目标水体分割结果确定模块，用于根据所述第一目标图像的所述第一水体分割结果，以及所述至少一个第二水体分割结果，确定所述第一目标图像的目标水体分割结果。
  12. 一种水体分割装置，包括：
    第三水体分割模块，用于将第二目标图像输入水体分割网络，得到所述第二目标图像的第三水体分割结果；
    第四水体分割模块，用于根据所述第二目标图像的光谱信息，对所述第二目标图像的水体区域进行分割，得到所述第二目标图像的第四水体分割结果；
    目标水体分割结果获取模块，用于将所述第二目标图像的所述第三水体分割结果和所述第四水体分割结果中共同包含的水体区域，作为所述第二目标图像的目标水体分割结果。
  13. 一种电子设备，其特征在于，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器被配置为调用所述存储器存储的指令，以执行权利要求1至10中任意一项所述的方法。
  14. 一种计算机可读存储介质，其上存储有计算机程序指令，其特征在于，所述计算机程序指令被处理器执行时实现权利要求1至10中任意一项所述的方法。
PCT/CN2021/103731 2020-10-13 2021-06-30 水体分割方法及装置、电子设备和存储介质 WO2022077958A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2022520222A JP2023502194A (ja) 2020-10-13 2021-06-30 水域分割方法及び装置、電子機器並びに記憶媒体

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011092192.5 2020-10-13
CN202011092192.5A CN112164083A (zh) 2020-10-13 2020-10-13 水体分割方法及装置、电子设备和存储介质

Publications (1)

Publication Number Publication Date
WO2022077958A1 true WO2022077958A1 (zh) 2022-04-21

Family

ID=73866763

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/103731 WO2022077958A1 (zh) 2020-10-13 2021-06-30 水体分割方法及装置、电子设备和存储介质

Country Status (3)

Country Link
JP (1) JP2023502194A (zh)
CN (1) CN112164083A (zh)
WO (1) WO2022077958A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115170947A (zh) * 2022-05-12 2022-10-11 广东省科学院广州地理研究所 基于遥感图像的河口浑浊带与水体分类方法、装置及设备
CN117409203A (zh) * 2023-11-14 2024-01-16 自然资源部国土卫星遥感应用中心 一种浅水湖泊面积提取的方法
CN117994797A (zh) * 2024-04-02 2024-05-07 杭州海康威视数字技术股份有限公司 一种水尺读数方法、装置、存储介质和电子设备

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112164083A (zh) * 2020-10-13 2021-01-01 上海商汤智能科技有限公司 水体分割方法及装置、电子设备和存储介质
CN112668530A (zh) * 2021-01-04 2021-04-16 北京简巨科技有限公司 水体提取方法、装置、电子设备及存储介质
CN113343945B (zh) * 2021-08-02 2021-12-31 航天宏图信息技术股份有限公司 水体识别方法、装置、电子设备及存储介质
CN114332637B (zh) * 2022-03-17 2022-08-30 北京航空航天大学杭州创新研究院 遥感影像水体提取方法、遥感影像水体提取的交互方法

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109977801A (zh) * 2019-03-08 2019-07-05 中国水利水电科学研究院 一种联合光学和雷达的区域水体快速动态提取方法及系统
CN111127486A (zh) * 2019-12-25 2020-05-08 Oppo广东移动通信有限公司 图像分割方法、装置、终端及存储介质
CN111160349A (zh) * 2019-12-16 2020-05-15 广州地理研究所 基于哨兵遥感数据的水体提取方法及装置、设备
CN112164083A (zh) * 2020-10-13 2021-01-01 上海商汤智能科技有限公司 水体分割方法及装置、电子设备和存储介质

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109977801A (zh) * 2019-03-08 2019-07-05 中国水利水电科学研究院 一种联合光学和雷达的区域水体快速动态提取方法及系统
CN111160349A (zh) * 2019-12-16 2020-05-15 广州地理研究所 基于哨兵遥感数据的水体提取方法及装置、设备
CN111127486A (zh) * 2019-12-25 2020-05-08 Oppo广东移动通信有限公司 图像分割方法、装置、终端及存储介质
CN112164083A (zh) * 2020-10-13 2021-01-01 上海商汤智能科技有限公司 水体分割方法及装置、电子设备和存储介质

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115170947A (zh) * 2022-05-12 2022-10-11 广东省科学院广州地理研究所 基于遥感图像的河口浑浊带与水体分类方法、装置及设备
CN115170947B (zh) * 2022-05-12 2023-04-18 广东省科学院广州地理研究所 基于遥感图像的河口浑浊带与水体分类方法、装置及设备
CN117409203A (zh) * 2023-11-14 2024-01-16 自然资源部国土卫星遥感应用中心 一种浅水湖泊面积提取的方法
CN117409203B (zh) * 2023-11-14 2024-04-02 自然资源部国土卫星遥感应用中心 一种浅水湖泊面积提取的方法
CN117994797A (zh) * 2024-04-02 2024-05-07 杭州海康威视数字技术股份有限公司 一种水尺读数方法、装置、存储介质和电子设备

Also Published As

Publication number Publication date
CN112164083A (zh) 2021-01-01
JP2023502194A (ja) 2023-01-23

Similar Documents

Publication Publication Date Title
WO2022077958A1 (zh) 水体分割方法及装置、电子设备和存储介质
US10169896B2 (en) Rebuilding images based on historical image data
EP3226204B1 (en) Method and apparatus for intelligently capturing image
TWI753348B (zh) 位姿確定方法、位姿確定裝置、電子設備和電腦可讀儲存媒介
US10025472B2 (en) Method and apparatus for displaying data regarding a device's traversal through a region
US9280852B2 (en) Augmented reality virtual guide system
WO2020134866A1 (zh) 关键点检测方法及装置、电子设备和存储介质
WO2020181728A1 (zh) 图像处理方法及装置、电子设备和存储介质
WO2016192325A1 (zh) 视频文件的标识处理方法及装置
WO2023103377A1 (zh) 标定方法及装置、电子设备、存储介质及计算机程序产品
WO2019205605A1 (zh) 人脸特征点的定位方法及装置
TWI749532B (zh) 一種定位方法及定位裝置、電子設備和電腦可讀儲存媒介
US20210168279A1 (en) Document image correction method and apparatus
CN113450459B (zh) 目标物的三维模型构建方法及装置
EP3933753A1 (en) Method for processing image, related device and storage medium
WO2022017140A1 (zh) 目标检测方法及装置、电子设备和存储介质
CN111339880A (zh) 一种目标检测方法及装置、电子设备和存储介质
WO2023142554A1 (zh) 计数方法及装置、电子设备、存储介质和计算机程序产品
WO2023155350A1 (zh) 一种人群定位方法及装置、电子设备和存储介质
WO2023230927A1 (zh) 图像处理方法、装置及可读存储介质
WO2019174606A1 (zh) 图像处理方法和终端
CN113722410A (zh) 车位信息管理方法、装置、存储介质及电子设备
CN116310663A (zh) 点云配准效果的测试方法、设备、存储介质及程序产品
CN114170242A (zh) 样本图像的生成方法、图像预测方法、装置、设备及介质
CN114511712A (zh) 一种目标对象确定方法及装置、电子设备和存储介质

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2022520222

Country of ref document: JP

Kind code of ref document: A

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21879006

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21879006

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 22.09.2023)

122 Ep: pct application non-entry in european phase

Ref document number: 21879006

Country of ref document: EP

Kind code of ref document: A1