River probability prediction model training and river center line extraction method, device, equipment, medium and program

Abstract

The embodiment of the application discloses a river channel probability prediction model training and river channel center line extraction method, device, equipment, medium and program, wherein the method comprises: acquiring target training sample data; wherein the target training sample data comprises RGB remote sensing image sample data, initial river network distance feature sample data and center line heat map sample data; training a machine learning model according to the RGB remote sensing image sample data, the initial river network distance feature sample data and the center line heat map sample data to obtain a target river channel probability prediction model. The technical scheme of the embodiment of the application can improve the accuracy of weak feature river channel center line extraction.

Description

Technical Field

[0001] The embodiments of the present invention relate to the fields of remote sensing image processing and artificial intelligence, and particularly to a method, apparatus, equipment, medium and program for training a river channel probability prediction model and extracting the river channel centerline. Background Technology

[0002] River network data is the core foundational geographic data for hydrological analysis and comprehensive watershed management. In practical operations, there is a general spatial offset between conventional initial river network data and the accurately identifiable river centerlines in remote sensing images. Directly using uncorrected initial river network data will lead to distortion in the determination of river spatial location, thereby significantly reducing the accuracy of subsequent data analysis and application results.

[0003] Currently, existing deep learning remote sensing technologies for river networks and water body extraction primarily rely on medium- to high-resolution multispectral data or radar data for river identification and extraction. These technologies mainly extract and identify river boundaries and river network contours by mining multi-dimensional auxiliary information such as near-infrared bands, short-wave infrared bands, various water body indices, time-series water body distribution frequencies, and radar backscattering characteristics. However, the applicability of these existing technologies is significantly limited, heavily reliant on multi-band remote sensing data sources. When the operational scenario only allows the acquisition of RGB (Red-Green-Blue) three-band visible light remote sensing images, the limited image band dimensions prevent the formation of stable and distinguishable spectral response characteristics in the river channel. This renders traditional multispectral water body index algorithms, multi-source time-series hydrological frequency analysis methods, and areal water body semantic segmentation unsuitable for this scenario, hindering accurate river network extraction and location correction under RGB remote sensing images. Summary of the Invention

[0004] This invention provides a method, apparatus, equipment, medium, and program for training a river channel probability prediction model and extracting the river channel centerline, which can improve the accuracy of extracting the weak feature river channel centerline.

[0005] According to one aspect of the present invention, a method for training a river channel probability prediction model is provided, comprising: Acquire target training sample data; wherein, the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data; The machine learning model is trained based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

[0006] According to another aspect of the present invention, a method for extracting the centerline of a river channel is provided, comprising: Acquire RGB remote sensing image data and initial river network data of the target river channel; The initial river network distance feature data of the target river channel is determined based on the RGB remote sensing image data and the initial river network data; The RGB remote sensing image data of the target river channel and the initial river network distance feature data are input into the target river channel probability prediction model to obtain the target centerline heat map of the target river channel. Within the second preset distance threshold range of the initial river network, the corrected river centerline of the target river is determined according to the target centerline heat map; The target river channel probability prediction model is obtained by training the river channel probability prediction model training method described in any embodiment of the present invention.

[0007] According to another aspect of the present invention, a training apparatus for a river channel probability prediction model is provided, comprising: The target training sample data acquisition module is used to acquire target training sample data; wherein, the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data; The river channel probability prediction model training module is used to train the machine learning model based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

[0008] According to another aspect of the present invention, a river centerline extraction device is provided, comprising: The data acquisition module is used to acquire RGB remote sensing image data and initial river network data of the target river channel; The initial river network distance feature data determination module is used to determine the initial river network distance feature data of the target river channel based on the RGB remote sensing image data and the initial river network data. The target centerline heat map determination module is used to input the RGB remote sensing image data of the target river channel and the initial river network distance feature data into the target river channel probability prediction model to obtain the target centerline heat map of the target river channel. The corrected river centerline extraction module is used to determine the corrected river centerline of the target river based on the target centerline heat map within a second preset distance threshold range of the initial river network. The target river channel probability prediction model is obtained by training the river channel probability prediction model training method described in any embodiment of the present invention.

[0009] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and a memory communicatively connected to said at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the training method for the river channel probability prediction model or the river channel centerline extraction method according to any embodiment of the present invention.

[0010] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions, the computer instructions being configured to cause a processor to execute and implement the training method of the river channel probability prediction model or the river channel centerline extraction method according to any embodiment of the present invention.

[0011] According to another aspect of the present invention, a computer program product is also provided, comprising a computer program that, when executed by a processor, implements the training method for the river channel probability prediction model or the river channel centerline extraction method described in any embodiment of the present invention.

[0012] This invention acquires target training sample data, including RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data. Further, a machine learning model is trained using the RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data to obtain a target river channel probability prediction model. After training the target river channel probability prediction model, RGB remote sensing image data and initial river network data of the target river channel are acquired, and the initial river network distance feature data of the target river channel is determined based on these data. Further, the RGB remote sensing image data and initial river network distance feature data of the target river channel are input into the target river channel probability prediction model to obtain a target centerline heatmap of the target river channel. Based on this, within a second preset distance threshold range of the initial river network, the corrected river channel centerline of the target river channel is determined based on the target centerline heatmap. This scheme solves the problem of weak river channel features and difficulty in correcting the river channel centerline under RGB remote sensing images, and can improve the accuracy of extracting the centerline of weak-feature rivers.

[0013] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0015] Figure 1 This is a flowchart of a training method for a river channel probability prediction model provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a training method for a river channel probability prediction model provided in Embodiment 2 of the present invention; Figure 3 This is a flowchart of a method for extracting the centerline of a river channel provided in Embodiment 3 of the present invention; Figure 4 This is a schematic diagram of the extraction and correction effect of the corrected river centerline of a target river provided in Embodiment 3 of the present invention; Figure 5 This is a schematic diagram of a training device for a river channel probability prediction model provided in Embodiment 4 of the present invention; Figure 6 This is a schematic diagram of a river centerline extraction device provided in Embodiment 5 of the present invention; Figure 7 This is a schematic diagram of the structure of an electronic device provided in Embodiment Six of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0017] It should be noted that the terms "first," "second," and "target," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0018] Example 1 Figure 1 This is a flowchart of a training method for a river channel probability prediction model provided in Embodiment 1 of the present invention. This embodiment is applicable to the training of a river channel probability prediction model based on RGB remote sensing images. The method can be executed by a training device for the river channel probability prediction model. This device can be implemented in software and / or hardware, and is generally integrated into an electronic device. This electronic device can be a terminal device or a server device, as long as it can execute the training method for the river channel probability prediction model. The present invention does not limit the specific type of electronic device. Correspondingly, as... Figure 1 As shown, the method includes the following operations: S110. Obtain target training sample data; wherein, the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data.

[0019] The target training sample data can be sample data used to train a machine learning model to obtain a target river channel probability prediction model. For example, the target training sample data may include, but is not limited to, RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data. This embodiment of the invention does not limit the specific content included in the target training sample data.

[0020] In this embodiment of the invention, to train a river channel probability prediction model for river channel identification, target training sample data can be constructed. Specifically, the acquired target training sample data may include, but is not limited to, RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data. The RGB remote sensing image sample data can be multiple sets of pre-collected remote sensing image data corresponding to different rivers, and this remote sensing image sample data only contains image data in the red, green, and blue bands. The initial river network distance feature sample data can be used to characterize the spatial proximity of each pixel position in the RGB remote sensing image sample data to the corresponding initial river network. The initial river network can be a river centerline or river network vector initially generated based on the RGB remote sensing image sample data and topographic data. The centerline heatmap sample data can be used to characterize the probability that each pixel position in the RGB remote sensing image sample data belongs to the area near the river centerline.

[0021] S120. The machine learning model is trained based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

[0022] The target river channel probability prediction model can be a trained machine learning model that can be used for river channel identification.

[0023] Accordingly, after acquiring the target training sample data, the RGB remote sensing image sample data and the initial river network distance feature sample data can be used as model input data, and the centerline heatmap sample data can be used as supervision labels to train the machine learning model, thereby obtaining a trained target river channel probability prediction model. During the training process, the machine learning model can learn the correspondence between the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heatmap sample data. This allows the target river channel probability prediction model to correct the river channel centerline position based on the RGB remote sensing image when there is a shift in the initial river network, thereby improving the automation, positional accuracy, and continuity of weak feature river channel centerline extraction. In a specific example, the machine learning model can be any deep learning model.

[0024] This invention acquires target training sample data, including RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data. Further, a machine learning model is trained based on the RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data to obtain a target river channel probability prediction model. This approach solves the problem of weak river channel features and difficulty in correcting river channel centerlines in RGB remote sensing images, thus improving the accuracy of extracting weak-feature river channel centerlines.

[0025] Example 2 Figure 2 This is a flowchart of a training method for a river channel probability prediction model provided in Embodiment 2 of the present invention. This embodiment is a specific embodiment based on the above embodiment. In this embodiment, when the target training sample data also includes river channel area labeled sample data, specific optional implementation methods are given for obtaining the target training sample data, and training the machine learning model based on RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data to obtain the target river channel probability prediction model. Correspondingly, as Figure 2 As shown, the method in this embodiment may include: S210. Acquire RGB remote sensing image sample data, initial river network sample data, and manually calibrated river centerline sample data.

[0026] The initial river network sample data can be river centerlines or river network vectors initially generated based on RGB remote sensing image sample data and topographic data. The manually calibrated river centerline sample data can be river centerline vector data manually delineated and calibrated for the river area corresponding to the RGB remote sensing image sample data.

[0027] In this embodiment of the invention, to train a river channel probability prediction model for river channel identification, multiple sets of RGB remote sensing image sample data corresponding to different rivers can first be acquired. For example, the RGB remote sensing image sample data can be sub-meter level RGB remote sensing images, i.e., spatial resolution better than 1m, containing only the three visible light bands: red (R), green (G), and blue (B). Further, initial river network sample data can be obtained based on the RGB remote sensing image sample data, combined with DEM (Digital Elevation Model) topographic analysis, historical river system data, or existing vector data. Simultaneously, river centerlines can be manually vectorized from the RGB remote sensing image sample data, existing river system vector data, or concurrent high-resolution reference images, and then manually calibrated river centerline sample data can be obtained after fracture repair, hanging line deletion, and topological connectivity checks.

[0028] Optionally, to ensure spatial consistency and matching accuracy of multi-source data, various types of sample data can be preprocessed by projection unification, spatial registration and region cropping before proceeding to the subsequent feature calculation and model training process.

[0029] S220. Determine the initial river network distance feature sample data based on the RGB remote sensing image sample data and the initial river network sample data.

[0030] Correspondingly, after acquiring RGB remote sensing image sample data and initial river network sample data, the spatial proximity between the position of each pixel in the RGB remote sensing image sample data and the initial river network sample data can be calculated based on the RGB remote sensing image sample data and the initial river network sample data, and used as the initial river network distance feature sample data.

[0031] Optionally, linear texture features can be calculated based on RGB remote sensing image sample data to enhance the response of elongated channels, valleys, and riverbed textures.

[0032] In an optional embodiment of the present invention, determining the initial river network distance feature sample data based on the RGB remote sensing image sample data and the initial river network sample data may include: The initial river network distance feature sample data is determined based on the RGB remote sensing image sample data and the initial river network sample data according to the following formula: ; in, The initial river network distance feature sample data, The position of each pixel in the RGB remote sensing image sample data Distance from the first target in the initial river network For distance attenuation scale, This represents the natural exponential function. The larger the value, the closer the cell is to the initial river network; The smaller the value, the farther the pixel is from the initial river network.

[0033] The first target distance can be the shortest distance between each pixel location in the RGB remote sensing image sample data and the initial river network.

[0034] In a specific example, if the pixel size of the RGB remote sensing image sample data is 1m, and the initial river network's planar offset relative to the true river centerline is typically no more than 10m, then the distance attenuation scale... The value can be 10 pixels. When the initial river network offset is large or the RGB remote sensing image resolution is low, it can be increased accordingly. For example, you can The number of pixels is increased to 15-20. In this embodiment of the invention, the specific value of the distance attenuation scale is not limited.

[0035] S230. Generate the centerline heat map sample data based on the RGB remote sensing image sample data and the artificially calibrated river centerline sample data.

[0036] Specifically, deep learning segmentation models are typically better suited to using area-based, rasterized supervision labels. Directly using linear, manually labeled river centerline samples as training labels results in a low percentage of effective positive samples and insufficient spatial tolerance, making it difficult for the model to fit the continuous probability distribution near the river channel. For these reasons, linear, manually labeled river centerline samples cannot be directly input into area-based segmentation models for training. Therefore, it is necessary to convert RGB remote sensing image sample data and manually labeled river centerline sample data to generate centerline heatmap sample data with area-based annotation characteristics. The above solution maintains the manual annotation work at the linear centerline level, reducing annotation costs while automatically generating area-based supervision labels, meeting the training requirements of deep learning models.

[0037] In an optional embodiment of the present invention, generating the centerline heatmap sample data based on the RGB remote sensing image sample data and the artificially calibrated river centerline sample data may include: The centerline heatmap sample data is generated based on the RGB remote sensing image sample data and the manually calibrated river centerline sample data using the following formula: ; ; in, The position of each pixel in the RGB remote sensing image sample data The probability of falling within a first preset distance threshold range of the artificially calibrated river centerline is determined based on the pixel positions in the RGB remote sensing image sample data. The probability of the sample data for the centerline heatmap being located within a first preset distance threshold range of the artificially calibrated river centerline is used to determine the centerline heatmap sample data. The pixel positions in the RGB remote sensing image sample data The second target distance between the artificially calibrated river centerline and the target distance. The attenuation scale of the heat map For any point on the center line of the artificially calibrated river, The artificially calibrated river centerline is used.

[0038] The first preset distance threshold can be a pre-defined distance range used to characterize the area adjacent to the river's centerline. The second target distance can be the location of each pixel in the RGB remote sensing image sample data. The shortest Euclidean distance between the river and the artificially calibrated centerline.

[0039] In a specific example, if the pixel size of the resampled RGB remote sensing image is 1m, then the heatmap attenuation scale... The value can be 3 pixels, so that the thermal map response at a position 3 pixels away from the artificially calibrated river centerline is about exp(-1 / 2), and the overall distribution shows that the thermal response value is lower the farther away from the artificially calibrated river centerline.

[0040] S240. Generate the river area annotation sample data based on the RGB remote sensing image sample data, the artificially calibrated river centerline sample data, and the preset buffer radius.

[0041] The preset buffer radius can be a pre-set radius value used to define the river channel range. The river channel area annotation sample data can be binary raster labels used to characterize whether a pixel location belongs to the river channel range.

[0042] Specifically, after acquiring RGB remote sensing image sample data and artificially calibrated river centerline sample data, the pixel positions in the RGB remote sensing image sample data that are within the preset buffer radius of the artificially calibrated river centerline can be marked as the river channel range, and the remaining pixel positions can be marked as the non-river channel range. The labels obtained after the above marking process are the river channel area annotation sample data.

[0043] It should be noted that the river area labeled sample data is only used as auxiliary supervision information during the model training phase. It is used to increase the spatial tolerance of the linear centerline label, help the model learn the differences between the river neighborhood and the non-river background, and thus improve the continuity and stability of the centerline heatmap.

[0044] S250. The machine learning model is trained based on the RGB remote sensing image sample data, the initial river network distance feature sample data, the centerline heat map sample data, and the river channel area annotation sample data to obtain the target river channel probability prediction model.

[0045] Accordingly, after acquiring RGB remote sensing image sample data, initial river network distance feature sample data, centerline heatmap sample data, and river channel area labeled sample data, the RGB remote sensing image sample data and initial river network distance feature sample data can be used as model input data, the centerline heatmap sample data as the primary supervision label, and the river channel area labeled sample data as the auxiliary label to train the machine learning model, thereby obtaining a target river channel probability prediction model for river channel identification. The initial river network distance feature sample data is used to provide candidate spatial ranges but is not the sole basis for localization; the model still needs to correct the river channel centerline position based on the riverbed texture, gully morphology, and slender linear cues on the RGB remote sensing images.

[0046] Optionally, a deep learning network with an encoder-decoder architecture can be used for model training. The model training process can simultaneously employ pixel classification loss and region overlap loss. This embodiment of the invention does not limit the specific network structure or the specific implementation of the loss function.

[0047] This invention acquires RGB remote sensing image sample data, initial river network sample data, and manually calibrated river centerline sample data, and then determines initial river network distance feature sample data based on the RGB remote sensing image sample data and the initial river network sample data. Simultaneously, it generates centerline heatmap sample data based on the RGB remote sensing image sample data and the manually calibrated river centerline sample data; and generates river channel area annotation sample data based on the RGB remote sensing image sample data, the manually calibrated river centerline sample data, and a preset buffer radius. Based on this, a machine learning model is trained using the RGB remote sensing image sample data, the initial river network distance feature sample data, the centerline heatmap sample data, and the river channel area annotation sample data to obtain a target river channel probability prediction model.

[0048] The target training sample data is acquired, including RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data. Further, a machine learning model is trained based on the RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data to obtain a target river channel probability prediction model. This approach addresses the problem of weak river channel features and difficulty in correcting river channel centerlines in RGB remote sensing images, improving the accuracy of centerline extraction for weak-feature rivers.

[0049] Example 3 Figure 3 This is a flowchart of a method for extracting the centerline of a river channel according to Embodiment 3 of the present invention. This embodiment is applicable to the extraction of the river channel centerline using a river channel probability prediction model trained based on RGB remote sensing images. This method can be executed by a river channel centerline extraction device, which can be implemented in software and / or hardware, and is generally integrated into an electronic device. This electronic device can be a terminal device or a server device, as long as it can execute the river channel centerline extraction method. The present invention does not limit the specific type of electronic device. Correspondingly, as... Figure 3 As shown, the method includes the following operations: S310. Acquire RGB remote sensing image data and initial river network data of the target river channel.

[0050] The target river channel can be the river area for which centerline extraction is to be performed. The RGB remote sensing image data can be an RGB remote sensing image acquired for the target river channel. The initial river network data can be the river centerline or river network vector obtained from DEM topographic analysis, historical drainage data, or existing vector data, corresponding to the RGB remote sensing image data of the target river channel.

[0051] In this embodiment of the invention, when extracting the centerline of a target river channel, RGB remote sensing image data of the target river channel can be acquired first. Furthermore, initial river network data can be obtained based on the RGB remote sensing image data of the target river channel, combined with DEM topographic analysis, historical river system data, or existing vector data.

[0052] S320. Determine the initial river network distance feature data of the target river channel based on the RGB remote sensing image data and the initial river network data.

[0053] Among them, the initial river network distance feature data can be used to characterize the spatial proximity between the position of each pixel in the RGB remote sensing image data and the corresponding initial river network.

[0054] Accordingly, after acquiring the RGB remote sensing image data and initial river network data of the target river channel, the spatial proximity of each pixel position in the RGB remote sensing image data to the corresponding initial river network can be calculated as the initial river network distance feature data of the target river channel.

[0055] S330. Input the RGB remote sensing image data of the target river channel and the initial river network distance feature data into the target river channel probability prediction model to obtain the target centerline heat map of the target river channel.

[0056] Among them, the target centerline heat map can be the centerline heat map corresponding to the target river channel.

[0057] Accordingly, after obtaining the RGB remote sensing image data and initial river network distance feature data of the target river channel, these data can be input into the target river channel probability prediction model. The target river channel probability prediction model can output a heatmap of the target river channel's centerline based on the RGB remote sensing image data and initial river network distance feature data. It should be noted that each pixel in the target centerline heatmap corresponds to a probability value, representing the likelihood that the pixel location belongs to the vicinity of the river channel centerline. The target river channel probability prediction model can be trained using the training method for river channel probability prediction models described in any embodiment of this invention.

[0058] S340. Within the second preset distance threshold range of the initial river network, determine the corrected river centerline of the target river based on the target centerline heat map.

[0059] The second preset distance threshold can be a pre-set value used to characterize the neighborhood range of the initial river network of the target river. The corrected river centerline can be the precise centerline corresponding to the target river, representing the river centerline result obtained after the target river has undergone correction processing.

[0060] Accordingly, after obtaining the target centerline heatmap of the target river based on the target river channel probability prediction model, the heatmap can be mapped back to the original geographic space. Specifically, the neighborhood of the initial river network can be determined with the initial river network as the center and a second preset distance threshold as the radius. Further, within the neighborhood of the initial river network, the corrected river centerline can be determined based on the probability value, path length, and directional continuity of each pixel position in the target centerline heatmap. Directional continuity refers to the directional change of adjacent centerline segments not exceeding a preset threshold, used to avoid unreasonable reversals of the centerline. Specifically, paths with high centerline probability, reasonable path length, and continuous directional change can be preferentially selected as the corrected river centerline. After obtaining the corrected river centerline, it can be used for river network updates, hydrological analysis, and remote sensing interpretation result processing.

[0061] Figure 4 This is a schematic diagram illustrating the extraction and correction effect of the corrected river centerline of a target river according to Embodiment 3 of the present invention. In a specific example, such as... Figure 4 As shown, the red curve represents the initial river network obtained based on terrain analysis, the black dashed line represents the artificial reference line, i.e., the artificially calibrated river centerline, and the blue curve represents the corrected river centerline obtained through the river centerline extraction method provided in this embodiment of the invention. A significant spatial offset between the initial river network and the RGB remote sensing image is visible. The river centerline extraction method provided in this embodiment of the invention can re-extract the river centerline along the river texture and riverbed morphology of the RGB remote sensing image. Figure 4 The average error of the river centerline before correction was approximately 59.47m, and the average error after correction was approximately 0.81m, representing an improvement of approximately 58.66m. This demonstrates that in local river sections where the initial river network offset is significant, this method can utilize the weak channel characteristics in RGB remote sensing images to correct the centerline to a position closer to the artificial reference line.

[0062] This invention acquires RGB remote sensing image data and initial river network data of a target river channel, and determines initial river network distance feature data of the target river channel based on the RGB remote sensing image data and the initial river network data. Further, the RGB remote sensing image data and initial river network distance feature data of the target river channel are input into a target river channel probabilistic prediction model to obtain a target centerline heatmap of the target river channel. Based on this, within a second preset distance threshold range of the initial river network, the corrected river channel centerline of the target river channel is determined according to the target centerline heatmap. This scheme solves the problem of weak river channel features and difficulty in correcting the river channel centerline under RGB remote sensing images, and can improve the accuracy of extracting the weak feature river channel centerline.

[0063] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information in this technical solution comply with relevant laws and regulations and do not violate public order and good morals.

[0064] It should be noted that all information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for display, data used for analysis, etc.) involved in this disclosure are information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data comply with the relevant laws, regulations and standards of the relevant regions.

[0065] It should be noted that any arrangement or combination of the technical features in the above embodiments also falls within the protection scope of this invention.

[0066] Example 4 Figure 5 This is a schematic diagram of a training device for a river channel probability prediction model provided in Embodiment 4 of the present invention, as shown below. Figure 5 As shown, the device includes: a target training sample data acquisition module 410 and a river channel probability prediction model training module 420, wherein: The target training sample data acquisition module 410 is used to acquire target training sample data; wherein, the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data.

[0067] The river channel probability prediction model training module 420 is used to train the machine learning model based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

[0068] This invention acquires target training sample data, including RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data. Further, a machine learning model is trained based on the RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heatmap sample data to obtain a target river channel probability prediction model. This approach solves the problem of weak river channel features and difficulty in correcting river channel centerlines in RGB remote sensing images, thus improving the accuracy of extracting weak-feature river channel centerlines.

[0069] Optionally, the target training sample data further includes river channel area labeled sample data; the target training sample data acquisition module 410 is specifically used for: acquiring the RGB remote sensing image sample data, initial river network sample data, and manually calibrated river centerline sample data; determining the initial river network distance feature sample data based on the RGB remote sensing image sample data and the initial river network sample data; generating the centerline heatmap sample data based on the RGB remote sensing image sample data and the manually calibrated river centerline sample data; and generating the river channel area labeled sample data based on the RGB remote sensing image sample data, the manually calibrated river centerline sample data, and a preset buffer radius.

[0070] Optionally, the river channel probability prediction model training module 420 is specifically used to: train the machine learning model based on the RGB remote sensing image sample data, the initial river network distance feature sample data, the centerline heat map sample data, and the river channel area annotation sample data.

[0071] Optionally, the target training sample data acquisition module 410 is further configured to: determine the initial river network distance feature sample data based on the RGB remote sensing image sample data and the initial river network sample data according to the following formula: ; in, The initial river network distance feature sample data, The position of each pixel in the RGB remote sensing image sample data Distance from the first target in the initial river network This is the distance attenuation scale.

[0072] Optionally, the target training sample data acquisition module 410 is further configured to: generate the centerline heatmap sample data based on the RGB remote sensing image sample data and the artificially calibrated river centerline sample data according to the following formula: ; ; in, The position of each pixel in the RGB remote sensing image sample data The probability of falling within a first preset distance threshold range of the artificially calibrated river centerline is determined based on the pixel positions in the RGB remote sensing image sample data. The probability of the sample data for the centerline heatmap being located within a first preset distance threshold range of the artificially calibrated river centerline is used to determine the centerline heatmap sample data. The pixel positions in the RGB remote sensing image sample data The second target distance between the artificially calibrated river centerline and the target distance. The attenuation scale of the heat map For any point on the center line of the artificially calibrated river, The artificially calibrated river centerline is used.

[0073] The training device for the above-described river channel probability prediction model can execute the training method for the river channel probability prediction model provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method. Technical details not described in detail in this embodiment can be found in the training method for the river channel probability prediction model provided in any embodiment of the present invention.

[0074] Since the above-described training device for the river probability prediction model is capable of executing the training method for the river probability prediction model in the embodiments of the present invention, those skilled in the art can understand the specific implementation and various variations of the training device for the river probability prediction model in this embodiment based on the training method for the river probability prediction model described in the embodiments of the present invention. Therefore, how the training device for the river probability prediction model implements the training method for the river probability prediction model in the embodiments of the present invention will not be described in detail here. Any device used by those skilled in the art to implement the training method for the river probability prediction model in the embodiments of the present invention falls within the scope of protection of this application.

[0075] Example 5 Figure 6 This is a schematic diagram of a river centerline extraction device provided in Embodiment 5 of the present invention, as shown below. Figure 6 As shown, the device includes: a data acquisition module 510, an initial river network distance feature data determination module 520, a target centerline heat map determination module 530, and a corrected river channel centerline extraction module 540, wherein: The data acquisition module 510 is used to acquire RGB remote sensing image data and initial river network data of the target river channel.

[0076] The initial river network distance feature data determination module 520 is used to determine the initial river network distance feature data of the target river channel based on the RGB remote sensing image data and the initial river network data.

[0077] The target centerline heat map determination module 530 is used to input the RGB remote sensing image data of the target river channel and the initial river network distance feature data into the target river channel probability prediction model to obtain the target centerline heat map of the target river channel.

[0078] The corrected river centerline extraction module 540 is used to determine the corrected river centerline of the target river based on the target centerline heat map within a second preset distance threshold range of the initial river network.

[0079] The target river channel probability prediction model is trained using the river channel probability prediction model training method described in any embodiment of the present invention.

[0080] This invention acquires RGB remote sensing image data and initial river network data of a target river channel, and determines initial river network distance feature data of the target river channel based on the RGB remote sensing image data and the initial river network data. Further, the RGB remote sensing image data and initial river network distance feature data of the target river channel are input into a target river channel probabilistic prediction model to obtain a target centerline heatmap of the target river channel. Based on this, within a second preset distance threshold range of the initial river network, the corrected river channel centerline of the target river channel is determined according to the target centerline heatmap. This scheme solves the problem of weak river channel features and difficulty in correcting the river channel centerline under RGB remote sensing images, and can improve the accuracy of extracting the weak feature river channel centerline.

[0081] The aforementioned river centerline extraction device can execute the river centerline extraction method provided in any embodiment of the present invention, and possesses the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in this embodiment can be found in the river centerline extraction method provided in any embodiment of the present invention.

[0082] Since the river centerline extraction device described above is an apparatus capable of executing the river centerline extraction method in the embodiments of the present invention, those skilled in the art can understand the specific implementation and various variations of the river centerline extraction device in this embodiment based on the river centerline extraction method described in the embodiments of the present invention. Therefore, how the river centerline extraction device implements the river centerline extraction method in the embodiments of the present invention will not be described in detail here. Any apparatus used by those skilled in the art to implement the river centerline extraction method in the embodiments of the present invention falls within the scope of protection of this application.

[0083] Example 6 Figure 7A schematic diagram of an electronic device 10, which can be used to implement embodiments of the present invention, is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0084] like Figure 7 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 can also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0085] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0086] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, central processing unit (CPU), graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, digital signal processors (DSPs), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as methods for training river channel probabilistic prediction models or methods for extracting river channel centerlines.

[0087] In some embodiments, the method for training a river channel probability prediction model or the method for extracting a river channel centerline can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for training a river channel probability prediction model or the method for extracting a river channel centerline described above can be performed. Alternatively, in other embodiments, processor 11 can be configured to perform the method for training a river channel probability prediction model or the method for extracting a river channel centerline by any other suitable means (e.g., by means of firmware).

[0088] Optionally, the training method for the river channel probability prediction model may include: acquiring target training sample data; wherein the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data; training the machine learning model based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

[0089] Optionally, the method for extracting the river centerline may include: acquiring RGB remote sensing image data and initial river network data of the target river; determining initial river network distance feature data of the target river based on the RGB remote sensing image data and the initial river network data; inputting the RGB remote sensing image data and the initial river network distance feature data of the target river into the target river probability prediction model to obtain a target centerline heatmap of the target river; and determining a corrected river centerline of the target river based on the target centerline heatmap within a second preset distance threshold range of the initial river network; wherein the target river probability prediction model is trained using the training method of the river probability prediction model described in any embodiment of the present invention.

[0090] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0091] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0092] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0093] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0094] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0095] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0096] This application also discloses a computer program product, which includes a computer program that, when executed by a processor, implements the training method for the river channel probability prediction model or the river channel centerline extraction method provided in any embodiment of this application. This program product shares the same inventive concept as the training method for the river channel probability prediction model or the river channel centerline extraction method disclosed in the embodiments of this application, and therefore will not be described in detail here.

[0097] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0098] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A training method for a river channel probability prediction model, characterized in that, include: Acquire target training sample data; wherein, the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data; The machine learning model is trained based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

2. The method according to claim 1, characterized in that, The target training sample data also includes labeled sample data of river areas; obtaining the target training sample data includes: Acquire the RGB remote sensing image sample data, initial river network sample data, and manually calibrated river centerline sample data; The initial river network distance feature sample data is determined based on the RGB remote sensing image sample data and the initial river network sample data; The centerline heatmap sample data is generated based on the RGB remote sensing image sample data and the manually calibrated river centerline sample data. The river channel area annotation sample data is generated based on the RGB remote sensing image sample data, the manually calibrated river centerline sample data, and the preset buffer radius; The step of training the machine learning model based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heatmap sample data includes: The machine learning model is trained based on the RGB remote sensing image sample data, the initial river network distance feature sample data, the centerline heat map sample data, and the river channel area annotation sample data.

3. The method according to claim 2, characterized in that, The step of determining the initial river network distance feature sample data based on the RGB remote sensing image sample data and the initial river network sample data includes: The initial river network distance feature sample data is determined based on the RGB remote sensing image sample data and the initial river network sample data according to the following formula: ； in, The initial river network distance feature sample data, The position of each pixel in the RGB remote sensing image sample data Distance from the first target in the initial river network This is the distance attenuation scale.

4. The method according to claim 2, characterized in that, The step of generating the centerline heatmap sample data based on the RGB remote sensing image sample data and the manually calibrated river centerline sample data includes: The centerline heatmap sample data is generated based on the RGB remote sensing image sample data and the manually calibrated river centerline sample data using the following formula: ； ； in, The position of each pixel in the RGB remote sensing image sample data The probability of falling within a first preset distance threshold range of the artificially calibrated river centerline is determined based on the pixel positions in the RGB remote sensing image sample data. The probability of the sample data for the centerline heatmap being located within a first preset distance threshold range of the artificially calibrated river centerline is used to determine the centerline heatmap sample data. The pixel positions in the RGB remote sensing image sample data The second target distance between the artificially calibrated river centerline and the target distance. The attenuation scale of the heat map For any point on the center line of the artificially calibrated river, The artificially calibrated river centerline is used.

5. A method for extracting the centerline of a river channel, characterized in that, include: Acquire RGB remote sensing image data and initial river network data of the target river channel; The initial river network distance feature data of the target river channel is determined based on the RGB remote sensing image data and the initial river network data; The RGB remote sensing image data of the target river channel and the initial river network distance feature data are input into the target river channel probability prediction model to obtain the target centerline heat map of the target river channel. Within the second preset distance threshold range of the initial river network, the corrected river centerline of the target river is determined according to the target centerline heat map; The target river channel probability prediction model is obtained by training the river channel probability prediction model according to any one of claims 1-4.

6. A training device for a river channel probability prediction model, characterized in that, include: The target training sample data acquisition module is used to acquire target training sample data; wherein, the target training sample data includes RGB remote sensing image sample data, initial river network distance feature sample data, and centerline heat map sample data; The river channel probability prediction model training module is used to train the machine learning model based on the RGB remote sensing image sample data, the initial river network distance feature sample data, and the centerline heat map sample data to obtain the target river channel probability prediction model.

7. A device for extracting the centerline of a river channel, characterized in that, include: The data acquisition module is used to acquire RGB remote sensing image data and initial river network data of the target river channel; The initial river network distance feature data determination module is used to determine the initial river network distance feature data of the target river channel based on the RGB remote sensing image data and the initial river network data. The target centerline heat map determination module is used to input the RGB remote sensing image data of the target river channel and the initial river network distance feature data into the target river channel probability prediction model to obtain the target centerline heat map of the target river channel. The corrected river centerline extraction module is used to determine the corrected river centerline of the target river based on the target centerline heat map within a second preset distance threshold range of the initial river network. The target river channel probability prediction model is obtained by training the river channel probability prediction model according to any one of claims 1-4.

8. An electronic device, characterized in that, The electronic device includes: At least one processor; and a memory communicatively connected to said at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the training method of the river channel probability prediction model according to any one of claims 1-4, or to perform the river channel centerline extraction method according to claim 5.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the training method of the river channel probability prediction model according to any one of claims 1-4, or to execute the river channel centerline extraction method according to claim 5.

10. A computer program product, characterized in that, Includes a computer program / instruction, wherein when the computer program / instruction is executed by a processor, it implements the training method of the river channel probability prediction model according to any one of claims 1-4, or performs the river channel centerline extraction method according to claim 5.

Images