Visual flow measurement-based river channel main slip line automatic identification method and system
By adopting an automatic identification method for the main channel line of a river based on visual flow measurement, the river boundary is identified and an enhanced texture image is generated. The shoreline residue is removed and the main channel line area is extracted. This solves the problems of insufficient identification accuracy and poor robustness in the existing technology and achieves stable and reliable identification of the main channel line of the river.
Patent Information
- Application Number
- CN202511401532.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-28
- Publication Date
- 2025-12-23
AI Technical Summary
Existing methods for identifying main channel lines in rivers suffer from insufficient accuracy and poor robustness, making it difficult to meet the needs of practical applications. In particular, the identification effect is poor under complex lighting conditions, and they rely on specialized equipment and high-quality labeled data.
By collecting river image data, identifying river boundaries and generating enhanced texture images, and combining texture feature extraction and water masking, shoreline remnants are removed, the main channel area is extracted, and existing monitoring equipment is used for processing to adapt to complex lighting conditions.
It achieves long-term, stable, and reliable visual recognition of the main channel of the river without relying on specialized equipment, improving processing efficiency, adapting to complex natural lighting conditions, and reducing human and material costs.
Smart Images
Figure CN121190987A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the fields of water flow recognition and image processing technology, and in particular to an automatic identification method and system for the main channel of a river based on visual flow measurement. Background Technology
[0002] The main channel line, which is the line connecting the points of maximum vertical average flow velocity at various cross-sections of the river, is an important hydrological element reflecting the core axis of river flow dynamics. Accurate identification of the main channel line is of great significance for river evolution analysis, navigation safety, water-related engineering construction, and flood control and disaster relief command.
[0003] Traditional methods for monitoring the main channel of a river mainly rely on contact measurements, such as using hydrological instruments like ADCP (Acoustic Doppler Current Profiler) and current meters for mobile or fixed-point measurements. While these methods offer high accuracy, they have inherent drawbacks such as long operation cycles, high manpower and material costs, and significant risks (especially during flood season), making it difficult to achieve frequent and continuous monitoring of large-scale river channels.
[0004] In recent years, with the development of computer vision and remote sensing technologies, vision-based non-contact flow measurement technology has become a research hotspot in the industry. This type of technology analyzes river surface images or video sequences to invert the surface velocity field of water flow, thus providing a new technical approach for identifying main flow lines. However, existing visual recognition methods still have many limitations, resulting in insufficient accuracy, robustness, and practicality of their recognition results, making it difficult to meet the needs of practical applications.
[0005] Existing technologies also perform edge detection, region segmentation, or texture analysis on single-frame images to identify water flow features. However, the calculated velocity field is easily affected by factors such as wind force and photogrammetric parameters, and directly extracting the main flow line based on a velocity magnitude threshold has a large error.
[0006] Existing machine learning / deep learning-based methods include research attempting to segment main channels directly from images using models such as convolutional neural networks (CNNs). While these methods perform well in certain scenarios, their performance is highly dependent on training with large amounts of high-quality labeled data. Labeling main channels requires deep hydrological expertise, and data acquisition is difficult. Summary of the Invention
[0007] This application provides a method and system for automatic identification of main channel lines in rivers based on visual flow measurement, which can achieve long-term stable and reliable visual identification of main channel lines in rivers without relying on professional equipment and adapting to complex natural lighting conditions.
[0008] This application proposes an automatic identification method for the main channel line of a river based on visual flow measurement, including: Collect image data of the target river channel; Based on the base image of the target river channel, the river channel boundary in the image data is identified, wherein the base image is determined by successive comparisons of previously acquired image data, and the base image includes river channel boundaries marked in the previously acquired image data. Based on the identified river channel boundaries, texture features are extracted, and an enhanced texture image is generated based on the extracted texture features; and a binarized water body mask is generated based on the river channel boundaries. The shoreline remnants are removed based on the enhanced texture image and the water mask, and the main flow line region is extracted based on the enhanced texture image with the shoreline remnants removed.
[0009] Optionally, identifying the river boundary in the image data based on the base image of the target river includes: Pixel segmentation and edge detection are performed on the base image to obtain the basic river channel boundary; Based on the collected image data, it is registered with the basic river channel boundary; Based on the registration results, precise edge detection is performed on the image data within a region at a specified pixel distance from the basic river boundary to identify the river boundary.
[0010] Optionally, the base image may also be updated in the following manner: Calculate the similarity between the current image data and the gray-level histogram within the river boundary in the current base image; If the similarity is higher than a preset threshold, the lighting conditions are determined to be similar, and the image data is selected as a candidate frame. The selected candidate frames are subjected to absolute value difference and thresholding to obtain a change region mask; The stable region is obtained by removing short-term changes in the water surface area and moving objects on the riverbank based on the change area mask. The base image is updated based on the stable region.
[0011] Optionally, based on the identified river channel boundaries, texture features are extracted, and an enhanced texture image is generated based on the extracted texture features, including: Calculate rotation-invariant LBP features for each pixel within the river channel boundary region to obtain micro-texture; and, The contrast and entropy of the gray-level co-occurrence matrix (GLCM) within a set window are calculated in different directions to describe the roughness and clutter of the macroscopic texture. The microtexture, GLCM contrast, and GLCM entropy are linearly weighted to generate an enhanced texture image, wherein the microtexture and GLCM contrast are assigned positive weights, and the GLCM entropy is assigned negative weights, in order to highlight the main flow line region in the enhanced texture image.
[0012] Optionally, generating a binarized water mask based on the river channel boundary includes: Traverse the set of connected pixels within the river boundary to obtain the water body outline, and determine the hierarchical relationship of the water body outline. Based on the hierarchical relationship, the area outside the water body outline is filled with white or black, and the area inside the water body outline is filled with black or white to generate a water body mask.
[0013] Optionally, removing shoreline remnants based on the enhanced texture image and the water mask includes: The enhanced texture image is multiplied with the generated water mask to remove residual pixels along the shoreline; Extracting the main curve with the required pixel width based on the enhanced texture image after removing shoreline remnants includes: binarizing the multiplied result image using an adaptive thresholding method, and after binarization, connecting breakpoints and smoothing the contour through a morphological closure operation. Skeleton extraction is performed on the binarized image obtained by morphological closure operation to obtain the centerline with a width of one pixel.
[0014] Optionally, skeleton extraction is performed on the binarized image after morphological closure to obtain a centerline with a single pixel width, which includes the following two iterations: For any pixel P, P has a value of 1. Its 8 neighboring pixels are defined and denoted as P9 (P1), P2, P3, P4, P5, P6, P7, P8 in clockwise order. Pixels that meet all of the following conditions are designated as to be deleted: Condition 1: 2≤B(P)≤6, where B(P) is the number of non-zero neighbor points in the 8-neighborhood of point P; Condition 2: A(P) = 1, where A(P) is the number of times it changes from 0 to 1 when P2, P3, P4, P5, P6, P7, P8, P9 are in the order of P2; Condition 3: P2 * P4 * P6 = 0; Condition 4: P4 * P6 * P8 = 0; Pixels that satisfy condition one, condition two, and the following condition are designated as to be deleted: Condition 5: P2*P4*P8=0; Condition 6: P2 * P6 * P8 = 0; In each sub-iteration, the values of all marked pixels are set to 0 to delete the corresponding pixels; Perform iterations until no pixels are removed in a complete iteration to obtain a centerline with a single pixel width; Based on the center point, trace the next connected pixel in the connected direction until another endpoint or branch point is reached.
[0015] Optionally, extracting the main flow line region based on the enhanced texture image after removing shoreline remnants further includes: Determine the width of the river channel at the river channel boundary; The distribution of multiple centerlines obtained through iteration is determined based on the river boundary along the river width direction; The area with the densest distribution of centerlines within the river channel boundary is defined as the main channel area at the corresponding location.
[0016] This application also proposes an automatic identification system for the main channel of a river based on visual flow measurement, including a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the steps of the aforementioned automatic identification method for the main channel of a river based on visual flow measurement.
[0017] The method of this application embodiment can identify the main channel of a river using existing monitoring equipment. The river visual processing flow designed in this application does not introduce complex calculations of large models, improves processing efficiency, and can adapt to complex natural lighting conditions, achieving long-term stable and reliable visual identification of the main channel of the river.
[0018] The above description is merely an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0019] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a schematic diagram of the basic process of the automatic identification method for the main channel of a river based on visual flow measurement in this embodiment. Detailed Implementation
[0020] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
[0021] This application proposes an automatic identification method for the main channel line of a river based on visual flow measurement, such as... Figure 1 As shown, it includes the following steps: In step 101, image data of the target river channel is acquired. In some embodiments, video data of the river channel can be acquired through river monitoring equipment, and the required image data can be extracted from the video data.
[0022] In step 102, based on the base image of the target river channel, the river channel boundary in the image data is identified. The base image is determined by successively comparing previously acquired image data, and includes river channel boundaries marked in the previously acquired image data. For example, in some examples, typical pixels in the river channel can be extracted, and the river channel boundary in the image data can be identified through continuity relationships. In this example, the base image is updated over time based on river characteristics; that is, the interval between the acquisition time of the base image and the current acquisition time is within a set range. This constantly updated base image ensures that the river channel's topographical features, such as river width and local islands within the river channel, remain within a certain range of change, thereby improving the river channel identification speed.
[0023] In step 103, texture features are extracted based on the identified river channel boundaries, and an enhanced texture image is generated based on the extracted texture features; and a binarized water mask is generated based on the river channel boundaries. In a specific example, the extracted texture features may include micro-texture features and macro-texture features to enhance the flow lines.
[0024] In step 104, shoreline remnants are removed based on the enhanced texture image and the water mask, and the main flow line region is extracted based on the enhanced texture image with shoreline remnants removed.
[0025] In a specific example, riverbank remnants are further removed after texture enhancement, thereby identifying the main channel area based on the determined channels. The method of this application embodiment can identify the main channel of a river using existing monitoring equipment. Through the river vision processing flow designed in this application, the complex calculations of large models are not introduced, improving processing efficiency. Furthermore, it can adapt to complex natural lighting conditions, achieving long-term stable and reliable visual recognition of the main channel of a river.
[0026] In some embodiments, identifying the river boundary in the image data based on a base image of the target river includes: The base image is segmented and edge detected to obtain the basic river channel boundary. In some examples, this step involves segmentation and edge detection, and after determining the basic river channel boundary, it is further registered with the acquired image data.
[0027] Based on the collected image data, it is registered with the basic river channel boundary; Based on the registration results, precise edge detection is performed on the image data within a specified pixel distance from the base river channel boundary to identify the river channel boundary. As in the previous example, with updates to the base image, the changes in the current river channel boundary are within a certain range compared to the base image. Therefore, this example does not require complex boundary detection; instead, it performs edge detection, such as Canny edge detection, only within a specified pixel distance from the base river channel boundary. For example, a smaller pixel distance is set during normal water periods and a larger pixel distance during high water periods, thus accurately determining the current river channel boundary through localized edge detection. In some examples, the determined current river channel boundary can also be used to update the next base image, thereby maintaining the continuity of the base image.
[0028] This example proposes a method for accurately updating a base image based on video frames contained in video data collected by monitoring equipment. In some embodiments, the method further includes updating the base image in the following manner: Calculate the similarity between the current image data and the gray-level histogram within the river boundary in the current base image, for example, using the Bhattacharyya coefficient, or other methods, which will not be elaborated here.
[0029] If the similarity is higher than a preset threshold, the lighting conditions are determined to be similar or consistent, and the image data is selected as a candidate frame.
[0030] The selected candidate frames are subjected to absolute value difference and thresholding to obtain a change region mask. The shape and location of the change region are determined by the obtained change region mask, thereby filtering out transient changes caused by water surface ripples and moving objects, and retaining long-term and stable change regions that may be caused by shoreline vegetation growth, siltation, etc.
[0031] By removing short-term changes in the water surface area and moving objects on the riverbank using the change area mask, a stable region is obtained, which can then be updated using a moving average model.
[0032] The base image is updated based on the stable regions. In the updated base image, previously identified riverbank boundaries can be marked, allowing for further alignment using stable riverbank regions during registration, thereby improving registration efficiency.
[0033] In some embodiments, texture feature extraction is performed based on the identified river channel boundaries, and an enhanced texture image is generated based on the extracted texture features, including: Calculate rotation-invariant LBP features for each pixel within the river channel boundary region to obtain micro-texture; and, The contrast and entropy of the gray-level co-occurrence matrix (GLCM) within a set window are calculated in different directions to describe the roughness and clutter of the macroscopic texture. This application proposes a method for identifying main flow lines by combining image processing to obtain texture features, which characterizes the significant difference between the texture of the main flow line region and the non-main flow line region under normal lighting conditions. Based on this, micro and macro textures are extracted.
[0034] Based on the acquired microtexture, GLCM contrast, and GLCM entropy, a linear weighting is applied to generate an enhanced texture image. The microtexture and GLCM contrast are assigned positive weights, while the GLCM entropy is assigned a negative weight, to highlight the main flow line region in the enhanced texture image. This application addresses the characteristics of rapid water flow and highly directional, high-contrast textures in the main flow line region. By configuring the weighting of the two textures, a linear weighting model can be used in some examples. = * + * - * ,in , , For predetermined weights, the sum of their absolute values is 1. This is a rotation-invariant LBP feature for pixels. For GLCM contrast, The entropy of GLCM is significantly enhanced by weighting high LBP values, high contrast, and low entropy values to highlight the main flow line region from the image data, while suppressing random disturbances such as wind and waves, thus effectively obtaining the accurate main flow line region.
[0035] In some embodiments, generating a binarized water mask based on the river boundary includes: The process involves traversing the set of connected pixels within the river channel boundary to obtain the water body contour and determine its hierarchical relationship. Unlike determining the river channel boundary itself, this step further identifies connected pixels within the boundary to eliminate potentially fixed objects such as islands and sandbars within the river channel. The maximum outer contour of the obtained water body contour represents the outer boundary of the river channel. In the case of sandbars and islands, their contours will form "inner holes," i.e., sub-level contours.
[0036] Based on the hierarchical relationship, areas outside the water body outline are filled with white or black, while areas inside the water body outline are filled with black or white to generate a water body mask. That is, it distinguishes between filling the area inside and outside the boundary with black or white respectively. For example, it can generate a precise binary mask where the main body of the river is white (water body), and islands and sandbars are black (non-water body).
[0037] In some embodiments, removing shoreline remnants based on an enhanced texture image and the water mask includes: The enhanced texture image is multiplied with the generated water mask to remove residual pixels along the shoreline; Extracting the main contour line with the required pixel width based on the enhanced texture image after removing shoreline remnants includes: binarizing the multiplied image using an adaptive thresholding method, and then connecting breakpoints and smoothing the contour after binarization using a morphological closure operation. In a specific example, a structuring element, such as a circular or elliptical structuring element with a radius of 3-5 pixels, can be used to first dilate the binarized boundary to bridge smaller breakpoint gaps, followed by an erosion operation to restore the approximate original shape of the boundary. In this way, a continuous, closed contour line is ensured without significantly changing the original position of the boundary.
[0038] Skeleton extraction is performed on the binarized image obtained by morphological closure operation to obtain the centerline with a width of one pixel.
[0039] In some embodiments, performing skeleton extraction on a binarized image of a morphological closure operation to obtain a centerline with a single pixel width includes performing the following two iterations: For any pixel P, P has a value of 1. Its 8 neighboring pixels are defined and denoted as P9 (P1), P2, P3, P4, P5, P6, P7, P8 in clockwise order. Pixels that meet all of the following conditions are designated as to be deleted: Condition 1: 2≤B(P)≤6, where B(P) is the number of non-zero neighbor points in the 8-neighborhood of point P; Condition 2: A(P) = 1, where A(P) is the number of times it changes from 0 to 1 when P2, P3, P4, P5, P6, P7, P8, P9 are in the order of P2; Condition 3: P2 * P4 * P6 = 0; Condition 4: P4 * P6 * P8 = 0; Pixels that satisfy condition one, condition two, and the following condition are designated as to be deleted: Condition 5: P2*P4*P8=0; Condition 6: P2 * P6 * P8 = 0; In each sub-iteration, the values of all marked pixels are set to 0 to delete the corresponding pixels.
[0040] Perform iterations until no pixels are removed in a complete iteration to obtain a centerline with a single pixel width.
[0041] Based on the center point, trace the next connected pixel in the connected direction until another endpoint or branch point is reached.
[0042] In some embodiments, extracting the main flow line region based on the enhanced texture image after removing shoreline remnants further includes: Determine the width of the river channel at the river channel boundary; The distribution of multiple centerlines obtained through iteration is determined based on the river boundary along the river width. According to the aforementioned steps, streamline regions, including the main channel line, may be extracted. Based on the principle that the main channel line region has rapid water flow and strong directional texture, while the region outside the main channel line has gradually slower water flow and weaker texture, this example further determines the distribution of the obtained centerlines along the river width.
[0043] The area with the densest distribution of centerlines within the river channel boundary is determined as the main flow line area at the corresponding location. Based on the previous example, a denser distribution indicates a higher water flow velocity, and a longer centerline indicates a stronger direction. The required main flow line areas are determined accordingly based on actual needs.
[0044] The method of this application designs a base image that is updated sequentially, obtains the river boundary through local edge detection, and effectively highlights the main channel area by designing texture enhancement. Furthermore, the center line of the main channel is obtained through simple iteration. The scheme of this application uses existing monitoring equipment, does not require a complex calculation process, and can quickly and stably determine the main channel of the monitored river, which is convenient for engineering applications.
[0045] This application also proposes an automatic identification system for the main channel of a river based on visual flow measurement, including a processor and a memory. The memory stores a computer program, which, when executed by the processor, implements the steps of the aforementioned automatic identification method for the main channel of a river based on visual flow measurement.
[0046] Furthermore, although exemplary embodiments have been described herein, their scope includes any and all embodiments based on this disclosure that have equivalent elements, modifications, omissions, combinations (e.g., schemes involving intersections of various embodiments), adaptations, or changes. They are not limited to the examples described in this specification or during the implementation of this application, and such examples are to be construed as non-exclusive.
[0047] The above description is intended to be illustrative and not restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. Other embodiments can be used by those skilled in the art when reading the above description.
[0048] The above embodiments are merely exemplary embodiments of this disclosure. Those skilled in the art can make various modifications or equivalent substitutions to this invention within the scope of the disclosure, and such modifications or equivalent substitutions should also be considered to fall within the protection scope of this invention.
Claims
1. A method for automatic identification of the main channel line of a river based on visual flow measurement, characterized in that, include: Collect image data of the target river channel; Based on the base image of the target river channel, the river channel boundary in the image data is identified, wherein the base image is determined by successive comparisons of previously acquired image data, and the base image includes river channel boundaries marked in the previously acquired image data. Based on the identified river channel boundaries, texture features are extracted, and an enhanced texture image is generated based on the extracted texture features. In addition, a binarized water mask is generated based on the river boundary; The shoreline remnants are removed based on the enhanced texture image and the water mask, and the main flow line region is extracted based on the enhanced texture image with the shoreline remnants removed.
2. The automatic identification method for main channel flow based on visual flow measurement as described in claim 1, characterized in that, Based on the base image of the target river channel, identifying the river channel boundary in the image data includes: Pixel segmentation and edge detection are performed on the base image to obtain the basic river channel boundary; Based on the collected image data, it is registered with the basic river channel boundary; Based on the registration results, precise edge detection is performed on the image data within a region at a specified pixel distance from the basic river boundary to identify the river boundary.
3. The automatic identification method for main channel flow based on visual flow measurement as described in claim 2, characterized in that, This also includes updating the base image in the following manner: Calculate the similarity between the current image data and the gray-level histogram within the river boundary in the current base image; If the similarity is higher than a preset threshold, the lighting conditions are determined to be similar, and the image data is selected as a candidate frame. The selected candidate frames are subjected to absolute value difference and thresholding to obtain a change region mask; The stable region is obtained by removing short-term changes in the water surface area and moving objects on the riverbank based on the change area mask. The base image is updated based on the stable region.
4. The automatic identification method for main channel flow based on visual flow measurement as described in claim 2, characterized in that, Based on the identified river channel boundaries, texture features are extracted, and an enhanced texture image is generated based on the extracted texture features, including: Calculate rotation-invariant LBP features for each pixel within the river channel boundary region to obtain micro-texture; and, The contrast and entropy of the gray-level co-occurrence matrix (GLCM) within a set window are calculated in different directions to describe the roughness and clutter of the macroscopic texture. The microtexture, GLCM contrast, and GLCM entropy are linearly weighted to generate an enhanced texture image, wherein the microtexture and GLCM contrast are assigned positive weights, and the GLCM entropy is assigned negative weights, in order to highlight the main flow line region in the enhanced texture image.
5. The automatic identification method for main channel flow based on visual flow measurement as described in claim 4, characterized in that, The generation of a binarized water mask based on the river boundary includes: Traverse the set of connected pixels within the river boundary to obtain the water body outline, and determine the hierarchical relationship of the water body outline. Based on the hierarchical relationship, the area outside the water body outline is filled with white or black, and the area inside the water body outline is filled with black or white to generate a water body mask.
6. The automatic identification method for main channel flow based on visual flow measurement as described in claim 5, characterized in that, Removing shoreline remnants based on enhanced texture images and the water mask includes: The enhanced texture image is multiplied with the generated water mask to remove residual pixels along the shoreline; Extracting the main curve with the required pixel width based on the enhanced texture image after removing shoreline remnants includes: binarizing the multiplied result image using an adaptive thresholding method, and after binarization, connecting breakpoints and smoothing the contour through a morphological closure operation. Skeleton extraction is performed on the binarized image obtained by morphological closure operation to obtain the centerline with a width of one pixel.
7. The automatic identification method for main channel flow based on visual flow measurement as described in claim 6, characterized in that, Skeleton extraction is performed on the binarized image after morphological closure operations to obtain a centerline with a single pixel width, which involves the following two iterations: For any pixel P, P has a value of 1. Its 8 neighboring pixels are defined and denoted as P9 (P1), P2, P3, P4, P5, P6, P7, P8 in clockwise order. Pixels that meet all of the following conditions are designated as to be deleted: Condition 1: 2≤B(P)≤6, where B(P) is the number of non-zero neighbor points in the 8-neighborhood of point P; Condition 2: A(P) = 1, where A(P) is the number of times it changes from 0 to 1 when P2, P3, P4, P5, P6, P7, P8, P9 are in the order of P2; Condition 3: P2 * P4 * P6 = 0; Condition 4: P4 * P6 * P8 = 0; Pixels that satisfy condition one, condition two, and the following condition are designated as to be deleted: Condition 5: P2*P4*P8=0; Condition 6: P2 * P6 * P8 = 0; In each sub-iteration, the values of all marked pixels are set to 0 to delete the corresponding pixels; Perform iterations until no pixels are removed in a complete iteration to obtain a centerline with a single pixel width; Based on the center point, trace the next connected pixel in the connected direction until another endpoint or branch point is reached.
8. The method for automatic identification of main channel lines in a river based on visual flow measurement as described in claim 7, characterized in that, Extracting the main flow line region from the enhanced texture image based on removing shoreline remnants also includes: Determine the width of the river channel at the river channel boundary; The distribution of multiple centerlines obtained through iteration is determined based on the river boundary along the river width direction; The area with the densest distribution of centerlines within the river channel boundary is defined as the main channel area at the corresponding location.
9. An automatic identification system for the main channel of a river based on visual flow measurement, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program, which, when executed by the processor, implements the steps of the automatic identification method for the main channel of a river based on visual flow measurement as described in any one of claims 1 to 8.