A method and system for quickly measuring the amount of rural river silt at low cost

Abstract

The embodiment of the application provides a method and system for low-cost and rapid measurement of rural river channel siltation amount, and belongs to the technical field of river channel monitoring. The method comprises the following steps: a UAV flies along the center of the river channel to obtain an orthographic image of the river channel, and pre-processes the orthographic image to obtain a pre-processed image; water area segmentation is performed on the pre-processed image to identify a water body range, and a shoreline feature point in the water body range is extracted to determine a river channel center line; a control section is screened according to the river channel center line to obtain a section layout list; according to the section layout list, the siltation thickness of each control section is measured in combination with the coordinates of the section position to obtain section measurement data of each control section; and according to the section measurement data, the siltation thickness of each control section is adaptively fitted based on an adaptive optimization model, and the siltation amount of the river channel is obtained in combination with integral operation. The application can realize low-cost, efficient and rapid measurement of the siltation amount of the rural river channel.

Description

Technical Field

[0001] This application relates to the field of river monitoring technology, and in particular to a low-cost and rapid method and system for measuring the siltation of rural rivers. Background Technology

[0002] Rural waterways are scattered, small in scale, and numerous, closely related to farmland irrigation, flood control and drainage, and rural life and ecology. They also experience significant seasonal water level fluctuations and are easily disturbed by human activities. Existing dredging solutions mainly include: the cross-sectional topography method, which calculates volume differences by comparing underwater topography at different times using cross-sections; this method requires a large number of cross-sections and relies on depth sounders and post-processing software; underwater sonar measurement technology, which uses unmanned vessels equipped with multibeam or singlebeam depth sounders to construct three-dimensional silt models; this equipment is expensive and prone to failure in shallow water or weedy environments; mechanical rod insertion measurement, which involves manually inserting a graduated rod into the silt to the hard bottom to read the thickness; this is purely manual, involves a huge amount of fieldwork, human error cannot be quantified and controlled, and quality cannot be assessed in real time on-site; and UAV lidar and remote sensing technology, which calculates siltation volume by generating high-precision digital elevation models; this equipment is extremely expensive, and data post-processing is complex. Existing dredging solutions are mostly applicable to major rivers. However, for rural waterways that are numerous, wide-ranging, shallow, densely vegetated, and have limited funding, they generally suffer from problems such as large workload, high equipment costs, difficulty in controlling random errors, poor environmental adaptability, and inability to quickly perform on-site calculations.

[0003] In summary, the technical problems existing in the relevant technologies need to be improved. Summary of the Invention

[0004] The main objective of this application is to propose a low-cost and rapid method and system for measuring the siltation volume of rural waterways, so as to achieve a low-cost and high-efficiency rapid measurement of the siltation volume of rural waterways.

[0005] To achieve the above objectives, one aspect of this application proposes a low-cost and rapid method for calculating the siltation volume of rural waterways, the method comprising the following steps: The orthophoto of the river is obtained by flying along the center of the river with a drone, and the orthophoto is preprocessed to obtain a preprocessed image. The preprocessed image is segmented to identify the water body range, and shoreline feature points within the water body range are extracted to determine the river centerline; Based on the centerline of the river channel, control sections are selected to obtain a list of section layouts; Based on the cross-section layout list, the siltation thickness of each control cross-section is measured in conjunction with the cross-section location coordinates to obtain the cross-section measurement data of each control cross-section. Based on the cross-sectional measurement data, the sedimentation thickness of each control cross-section is adaptively fitted using an adaptive optimization model, and the sedimentation amount of the river channel is obtained by combining integral calculations.

[0006] In some embodiments, the preprocessed image is segmented to identify the water body range, and shoreline feature points within the water body range are extracted to determine the river centerline, including: Extract multi-channel features from the preprocessed image to construct a water body feature discrimination dataset; Based on the water body feature discrimination dataset, the preprocessed image is classified pixel-by-pixel using a lightweight semantic segmentation network, and morphological constraints are added to obtain a water body mask. Extract the closed outer contour of the water body mask to obtain the water body range of the entire river channel; Extract feature points from both sides of the shoreline within the water body to obtain a set of feature points on both banks; Along the direction of the river flow, the feature points in the feature point set on both banks are paired one by one according to the transverse direction of the river, forming multiple sets of transverse lines that cross the river. Calculate the spatial midpoint coordinates of the two feature points of the transverse line, and fit the spatial midpoint coordinates based on cubic spline interpolation to obtain the river centerline.

[0007] In some embodiments, the lightweight semantic segmentation network is a lightweight encoder-decoder symmetric semantic segmentation network. The encoder backbone adopts an improved MobileNetV2 lightweight inverse residual structure as a feature extractor, and the decoder adopts a skip connection, channel concatenation, and lightweight fusion convolution structure to restore the image resolution layer by layer. The output pixel segmentation probability map is activated by the Sigmoid function.

[0008] In some embodiments, a list of control sections is obtained by screening control sections based on the river centerline, including: Using the starting point of the river centerline as the origin of the coordinate system, cross sections perpendicular to the direction of the river are generated at fixed intervals along the longitudinal direction of the water flow on the river centerline, serving as basic equidistant candidate cross sections. The changes in river morphology are determined segment by segment along the centerline of the river channel. If any judgment condition is met, a new control section is added. The basic equidistant candidate sections and newly added control sections are deduplicated and verified, and the number of measuring points is allocated to generate a section layout list.

[0009] In some embodiments, the determination criteria include determination of bend curvature, determination of sudden changes in river width, determination of tributary confluence, and determination of water-related structures.

[0010] In some embodiments, a portable thickness measuring device is used to measure the silt thickness at each control section. The portable thickness measuring device integrates a damping module, an tilt sensing module, a laser ranging module, and a Bluetooth module. The damping module is used to control the penetration resistance, allowing the portable thickness measuring device to penetrate the silt at a uniform speed. The tilt sensing module is used to monitor the verticality of the rod in real time, allowing the portable thickness measuring device to penetrate the silt vertically. The laser ranging module is used to read the measurement data and upload it to a mobile terminal in real time via the Bluetooth module.

[0011] In some embodiments, based on the cross-sectional measurement data, an adaptive optimization model is used to adaptively fit the sedimentation thickness of each control cross-section, and the sedimentation amount of the river channel is obtained by combining integral calculations, including: The cross-sectional measurement data of each control section were fitted using a weighted parabolic model to generate the cross-sectional sedimentation thickness curves of each control section. The cross-sectional sedimentation thickness curve is subjected to lateral numerical integration to obtain the cross-sectional sedimentation area corresponding to each control section. Using the longitudinal mileage of each control section as the independent variable, independent non-uniform cubic spline interpolation functions are constructed for the model parameters, section width, and cross-sectional siltation area, respectively, to generate the cross-sectional siltation area curve at any longitudinal position of the river channel. The total amount of sedimentation is calculated using the adaptive composite Simpson integral based on the cross-sectional sedimentation area curve.

[0012] In some embodiments, an adaptive encryption mechanism is introduced when calculating the total sedimentation volume. The convergence of the results is judged by integrating the river channel in segments and calculating the relative error between the total sedimentation volume before and after encryption. The converged value is then rounded to obtain the total sedimentation volume of the entire river channel.

[0013] To achieve the above objectives, another aspect of this application proposes a low-cost and rapid system for measuring siltation in rural waterways, the system comprising: The image acquisition module is used to acquire orthophotos of the river channel by flying along the center of the river channel with a drone, and to preprocess the orthophotos to obtain a preprocessed image. The cross-section feature recognition module is used to segment the water area of ​​the preprocessed image to identify the water body range and extract shoreline feature points within the water body range to determine the river centerline. The cross-section screening module is used to screen control cross-sections based on the river centerline to obtain a cross-section layout list; The cross-section measurement module is used to measure the siltation thickness of each control section according to the cross-section layout list and the cross-section location coordinates, so as to obtain the cross-section measurement data of each control section. The siltation calculation module is used to adaptively fit the siltation thickness of each control section based on the cross-sectional measurement data and an adaptive optimization model, and then combine the integral calculation to obtain the siltation volume of the river channel.

[0014] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0015] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0016] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0017] The embodiments of this application include at least the following beneficial effects: This application provides a low-cost and rapid method and system for measuring the siltation of rural river channels. This scheme acquires orthophotos of the river channel through UAV flight and performs preprocessing. Compared with full-area grid aerial photography, it can significantly reduce the number of images collected, shorten the field aerial photography time, adapt to narrow and small rural river channels, and eliminate interference from water surface reflection, tree shadow distortion, segmented color difference, and lens distortion through preprocessing, outputting standardized images with a unified geographic coordinate system. Then, pixel-level water body recognition is achieved based on a lightweight semantic segmentation network, combined with morphological optimization, to accurately distinguish the main river channel water body, scattered water accumulation on the banks, and duckweed and shallows, completely locking the river water boundary and avoiding interference from non-target water areas. By extracting the inflection points and abrupt changes in width of the shorelines on both banks, and fitting the midpoints of paired shorelines to generate a smooth and continuous centerline, the true direction of the river channel is accurately restored. Then, control sections are evenly distributed in straight river sections. Control sections are automatically added in areas prone to siltation, such as sharp bends, abrupt changes in river width, tributary confluences, and bridge / dam areas. The number of measuring points is allocated according to the section type to balance survey workload and measurement accuracy, outputting a structured section layout list. This list is then directly distributed to drones for measuring point navigation, solving the problems of traditional manual point placement lacking precise coordinates and difficulty in locating sections in the field. Drone navigation locates the control section measuring points, and a lightweight thickness measuring device is used to measure siltation thickness. The measurement data is reproducible and highly reliable. An adaptive optimization model is used to fit the measurement data, and integral calculations are combined to obtain the river's siltation volume. From aerial map generation, automatic point placement, high-precision field measurement to automatic calculation of siltation volume across the entire area, the entire process significantly reduces the workload of manual surveying, mapping, recording, and indoor calculation. The entire solution can be completed using only low-cost drones, portable measuring rods, and ordinary mobile terminals. It is suitable for rural shallow water, aquatic plants, and narrow river conditions, and its measurement accuracy is superior to traditional manual cross-section survey methods, greatly shortening the river siltation survey cycle. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating a low-cost and rapid method for calculating siltation in rural waterways, as provided in an embodiment of this application. Figure 2 This is a schematic diagram of the overall structure of the portable thickness measuring device according to an embodiment of this application; Figure 3 This is a structural diagram of a low-cost, rapid system for calculating siltation in rural waterways, provided in an embodiment of this application. Figure 4 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application.

[0019] The diagram shows: 1. Portable damping penetration handle; 2. Rotary damping adjuster; 3. Measuring rod; 4. Depth scale; 5. External fixed scale rod; 6. Bottom conical drill bit. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0021] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0022] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0023] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0024] Before providing a detailed description of the embodiments of this application, some of the nouns and terms used in the embodiments of this application will be explained first. The nouns and terms used in the embodiments of this application shall be interpreted as follows: Silt deposition: This refers to the total volume or mass of sediments such as mud, silt, and debris deposited within a specified body of water (river, reservoir, ditch, pipeline, etc.) over a certain period of time. In the context of rural river channels, it specifically refers to the total volume of newly deposited silt layers on top of the existing hard base of the riverbed.

[0025] Cross-section: refers to a vertical section that is perpendicular to the direction of water flow in a river, spans the entire river channel, and extends from the left bank to the right bank.

[0026] Existing dredging methods are mostly applicable to large rivers. When applied to rural waterways that are numerous, widespread, shallow, densely vegetationed, and where funds are limited, problems arise such as large workload, high equipment costs, difficulty in controlling random errors, poor environmental adaptability, and inability to quickly measure siltation on-site. Therefore, this application provides a low-cost, rapid method and system for measuring siltation in rural waterways, enabling low-cost, high-efficiency, and rapid measurement of siltation levels.

[0027] This application provides a low-cost and rapid method and system for calculating siltation in rural river channels, relating to the field of river monitoring. This method and system can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle-mounted terminal, but is not limited to these. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the low-cost and rapid method and system for calculating siltation in rural river channels, but is not limited to the above forms.

[0028] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0029] Figure 1 This is an optional flowchart of a low-cost and rapid method for measuring siltation in rural waterways, provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S1 to S6: S1, the orthophoto of the river is obtained by flying the drone along the center of the river, and the orthophoto is preprocessed to obtain a preprocessed image. The starting and ending coordinates of the target rural river, along with its maximum width, direction, and coordinates of surrounding obstacles, are imported into the mobile terminal. The drone is set to fly along a parallel route along the river's longitudinal center, with its flight altitude set according to the river's width (e.g., 50m). A high-definition visible light gimbal camera is mounted on the drone, operating in a vertically downward orthogonal shooting mode with a fixed gimbal tilt of 90° to eliminate perspective distortion caused by tilted shooting. The flight speed is set to a constant 1.5–2.5 m / s to ensure high pixel resolution per image and accurately identify narrow riverbanks, aquatic plants, and shallow boundaries. The drone automatically cruises along the center of the target rural river, triggering the camera at equal intervals to capture images. It records the GPS coordinates, altitude, and timestamp of each image throughout the entire process. After completely covering the target rural river, the drone automatically returns, exporting the entire raw aerial image dataset. This raw image dataset is then preprocessed, including radiometric correction, geometric distortion correction, image color and illumination equalization, image stitching, noise filtering, coordinate registration, and cropping.

[0030] Specifically, the camera's built-in radiometric calibration parameters are read to correct the pixel grayscale values ​​of each original image. Highlight suppression is applied to the bright areas of the river surface to reduce white reflective patches on the water surface. The grayscale differences between the water body and the silt and shallows are distinguished, and the brightness and contrast benchmarks of all images across the entire river section are unified. Color differences in images taken in the early morning / late evening, on cloudy days, and on sunny days are eliminated to remove pixel brightness deviations caused by uneven lighting, strong light reflections from the water surface, tree shadows, and color differences from the camera sensor. Lens distortion correction is performed using the camera's intrinsic parameters (focal length, distortion coefficient) to correct image edge stretching and curvature. Combined with the GPS elevation and attitude data (pitch / roll / yaw angles) attached to each image, orthorectification is performed using a Digital Surface Model (DSM) to project the tilted pixels onto a horizontal ground coordinate system, ensuring that the image is a standard vertical orthophoto view to restore the true vertical plane. Extract common pixel features from all overlapping areas of the images, calculate the color balance matrix, adjust the RGB channel pixel values ​​one by one, eliminate the light and dark boundary lines at the stitching of adjacent images, and ensure a uniform color transition across the entire river channel.

[0031] Based on image feature point matching algorithms such as SIFT, feature points of shorelines, trees, and embankments in overlapping areas of adjacent images are matched to generate a complete and continuous single large map of the entire river channel, fully covering the entire water area and shoreline from the river's starting point to its ending point, ensuring no duplication or blank areas. The river imagery contains aquatic plant textures, ripples, and sensor salt-and-pepper noise; bilateral filtering is used to smooth only random noise in the water surface and mud areas, preserving key contour features such as shorelines, river edges, and silt boundaries. The stitched large image is bound to a unified WGS84 planar geographic coordinate system, assigning precise latitude and longitude coordinates to each pixel. The image is cropped according to the starting and ending boundaries of the target rural river channel, removing irrelevant farmland and open spaces on both sides of the river, retaining only the effective image containing the river water, and outputting a standardized preprocessed image (TIFF format, with geographic coordinate information).

[0032] S2, perform water area segmentation on the preprocessed image to identify the water body range, and extract shoreline feature points within the water body range to determine the river centerline; In some embodiments, the preprocessed image is segmented to identify the water body range, and shoreline feature points within the water body range are extracted to determine the river centerline, including the following steps: S21, extract the multi-channel features of the preprocessed image and construct a water body feature discrimination dataset; S22, Based on the water body feature discrimination dataset, perform pixel-by-pixel binary classification on the preprocessed image using a lightweight semantic segmentation network, and add morphological constraints to obtain a water body mask; S23, extract the closed outer contour of the water body mask to obtain the water body range of the entire river channel; S24, extract the feature points of the shoreline on both sides of the water body to obtain the set of feature points on both sides; S25, along the direction of the river flow, the feature points in the feature point set on both banks are paired one by one according to the transverse direction of the river, forming multiple sets of transverse lines that cross the river. S26. Calculate the spatial midpoint coordinates of the two feature points of the transverse connection, and fit the spatial midpoint coordinates based on cubic spline interpolation to obtain the river centerline.

[0033] Specifically, since pixels in water areas generally exhibit low brightness, low texture, and bluish-gray tones, while riverbank mud, weeds, and roads have high gradient and multicolored texture features, the RGB visible light channels, grayscale gradients, and texture features of the preprocessed image are extracted to construct a water feature discrimination dataset to distinguish four easily confused regions: shallow muddy water areas, deep water, dried-up mudflats, and riverbank vegetation. Then, using the water feature discrimination dataset as input and water and non-water pixels as classification labels, a lightweight semantic segmentation network performs pixel-by-pixel binary classification on the entire preprocessed image to obtain a preliminary segmented binary map of the water body. To address interference from surface ripples, duckweed, and silt in rural waterways, morphological constraints are added to eliminate missegmentation of scattered noise points on the water surface and missed segmentation of voids caused by weeds on the banks. Specifically, closing and opening operations are performed on the initially segmented binary water image. The closing operation fills in voids caused by duckweed and aquatic plants, while the opening operation removes non-target water body interference such as scattered puddles and small ditches on the banks, resulting in a continuous and complete mask map of the connected waterway region, i.e., the water mask. The closed outer contour of the water mask is then extracted to delineate the boundary region between the river water area and the land on both sides of the image; this contour represents the entire water area of ​​the river.

[0034] Subsequently, the direction was initially determined using the centerline of the water body mask. The water body contour was then divided into two independent closed edges: the left bank contour curve and the right bank contour curve, with the pixel sequences of both sides stored separately. Ordinary shoreline feature points were uniformly sampled on both the left and right bank contour curves, and key feature inflection points were extracted as shoreline feature points. Required feature points included river bends, abrupt changes in river width, tributary confluence points, and bridge crossing endpoints. For smooth, straight river sections, ordinary shoreline feature points were uniformly sampled at fixed intervals (e.g., 5 meters of actual distance). All shoreline feature points were simultaneously bound to image geographic coordinates, and the pixel position and corresponding actual latitude and longitude were recorded. False feature points caused by weeds and rocks were removed, retaining only continuous, smooth shoreline feature points that accurately reflect the river's direction. This resulted in the left bank feature point set and the right bank feature point set, which together constitute the feature point sets for both banks.

[0035] Along the river's flow direction, feature points on the left and right banks are paired one-to-one according to the river's transverse direction. Each pair of paired points is perpendicular to the local river direction, forming multiple transverse lines spanning the river. For each pair of left and right bank feature points, the spatial midpoint coordinates of the line connecting the two points are calculated. All midpoints constitute the original discrete point set of the centerline. Based on cubic spline interpolation, curve fitting is performed on the original discrete point set of the centerline to eliminate jagged edges and offsets caused by sampling point jitter. The fitted result is a smooth, continuous single curve that runs through the entire river channel, which is the final river centerline. The river centerline is discretized into a vector curve with coordinates, and the latitude and longitude of each node and the curve direction parameters are stored as the core reference data for setting up the control sections in step S3.

[0036] In some embodiments, the lightweight semantic segmentation network is a lightweight encoder-decoder symmetric semantic segmentation network. The encoder backbone adopts an improved MobileNetV2 lightweight inverse residual structure as a feature extractor, and the decoder adopts a skip connection, channel concatenation, and lightweight fusion convolution structure to restore the image resolution layer by layer. The output pixel segmentation probability map is activated by the Sigmoid function.

[0037] Specifically, in the encoder, the shallow feature layer extracts the underlying RGB texture, edges, and brightness gradient features of the image by downsampling by 2 times using a single-layer standard 3×3 convolution. This is used to capture details of the shoreline, water ripples, and duckweed boundaries. Subsequently, two sets of inverse residual modules are stacked with a dilation coefficient t=1, performing only channel upscaling and preserving shallow details without depth compression to avoid losing small contours of the water body. The middle feature layer stacks three sets of inverse residual modules with a dilation coefficient t=6, continuously downsampling with a stride of 2 to extract mesoscale semantic features, distinguishing the color and texture differences of large continuous water surfaces, shore vegetation, and tidal flat silt. A lightweight attention module (SE) is embedded to weight and enhance the blue channel features of the water body, suppressing redundant feature interference from shore mud and vegetation. The deep feature layer consists of one set of highly dilated inverse residual modules with a dilation coefficient t=6, no downsampling with a stride of 1, outputting a global semantic feature map. Dilated convolution is used to expand the receptive field, enabling the identification of long-distance continuous river water bodies in one go, solving the problem of segmentation of narrow river segments. The encoder outputs three types of feature maps: shallow fine-grained features (1 / 4 of the original image size), which preserves the shoreline and small boundaries of aquatic plants; mid-level transitional features (1 / 8 of the original image size), which balances details and semantics; and deep global semantic features (1 / 16 of the original image size), which distinguishes the overall water body region.

[0038] The decoder employs a skip connection, channel concatenation, and a 1×1 lightweight fusion convolutional structure to restore image resolution layer by layer, achieving pixel-level segmentation. Deep global features are upsampled by 2x bilinearly and concatenated with mid-layer feature channels. A 1×1 convolution is used to compress redundant channels, achieving fusion of global water body regions and local mudflat features. The fused features are upsampled again by 2x and concatenated with the encoder's shallow fine-grained features. These shallow features accurately complete river edges, duckweed gaps, and shallow mudflat boundaries, addressing the contour blurring issue common in lightweight networks. The fused features undergo two 3×3 depthwise separable convolutions for dimensionality reduction, and are then mapped to a 2-channel feature map via a 1×1 convolution, corresponding to two pixel labels: channel 0 for non-water bodies and channel 1 for river water bodies. The output pixel segmentation probability map is activated by the Sigmoid function, resulting in a binary water body map.

[0039] For rural river scenarios, a weighted binary classification Dice loss and cross-entropy hybrid loss function is used. Since the proportion of water in rural rivers is much lower than that of land, a water pixel loss weight coefficient of 1.8 is set to alleviate sample imbalance and avoid classifying large areas of farmland and riverbanks as the main body. To address water hole formation caused by duckweed, a morphological constraint loss is added to penalize voids within the segmentation mask. The network outputs a probability threshold of 0.5 to obtain a binary water mask, which directly integrates with morphological opening / closing operations without additional format conversion. The training set consists entirely of orthophotos of rural river areas taken by UAVs, covering typical conditions such as shallow water and silt, water surface reflection, duckweed coverage, weed obstruction, wide and narrow bends, and tributary confluences, enhancing the network's ability to generalize and recognize complex river environments.

[0040] S3. Based on the centerline of the river channel, select control sections to obtain a list of section layouts; In some embodiments, a list of control sections is obtained by screening control sections based on the river centerline, including the following steps: S31, with the starting point of the river centerline as the origin of the coordinate system, generates cross sections perpendicular to the direction of the river along the longitudinal direction of the water flow at fixed intervals on the river centerline, which serve as the basic equidistant candidate cross sections. S32, traverse the river centerline segment by segment to determine the changes in river morphology, and add a new control section if any judgment condition is met; S33 performs deduplication verification on the basic equidistant candidate cross sections and newly added control cross sections, allocates the number of measuring points, and generates a cross section layout list.

[0041] Specifically, taking the starting point of the river centerline as the coordinate origin, the distance *s* along the longitudinal direction of the water flow is calculated for each point. At fixed intervals ΔL (default 500m-2000m, automatically recommended by the system based on the total river length), cross-sections perpendicular to the river's direction are generated at the centerline every ΔL distance. These cross-sections serve as basic equidistant candidate cross-sections (i.e., control cross-sections). The system automatically records the distance along the river, the coordinates of the cross-section centerline, and the initial cross-section number for each candidate cross-section. If the difference in distance between the basic equidistant candidate cross-section and the newly added control cross-section is less than 30m, it is considered a cross-section duplication, and only one cross-section is retained to avoid redundant measuring points. The number of measuring points is assigned to each control cross-section according to its type. For example, 5-6 measuring points are assigned to ordinary straight equidistant cross-sections, while 7 measuring points are assigned to cross-sections with bends, abrupt width changes, tributary confluences, and bridges. For cross-sections with complex siltation morphologies, increasing the number of measuring points improves the fitting accuracy. Integrate all the filtered cross-section information to form a structured cross-section layout list. Each cross-section in the list contains the following fields: cross-section number, longitudinal distance along the river starting point (i.e., distance s along the river), reason for layout (equidistant / bend / sudden width change / bridge / tributary confluence), geographical coordinates of the cross-section center, river width at the cross-section, total number of cross-section measuring points and lateral distribution of measuring points (left bank, equally divided points, right bank).

[0042] In some embodiments, the determination criteria include determination of bend curvature, determination of sudden changes in river width, determination of tributary confluence, and determination of water-related structures.

[0043] Specifically, the local curvature of the river channel is calculated point by point. When the local curvature exceeds a preset curvature threshold (default 0.05m⁻¹), it is identified as a sharp bend. A control section is added at the apex of the bend. The sharper the bend and the smaller the arc, the greater the curvature value. The width of adjacent river sections is continuously calculated. When the rate of change of width between adjacent sections exceeds a preset rate of change threshold (default 0.3), it indicates a sudden narrowing / widening of the river channel. A control section is added at the location of this abrupt change. The confluence of the main channel and tributaries is identified, and a control section is added downstream of the tributary confluence, such as 20m downstream. Cross-river structures in the pre-processed image, such as bridges, dams, and culverts, are identified, and control sections are added upstream and downstream of these structures, such as 5m upstream and 10m downstream.

[0044] S4. Based on the cross-section layout list, the siltation thickness of each control cross-section is measured in conjunction with the cross-section location coordinates to obtain the cross-section measurement data of each control cross-section. The operator loads the cross-section layout list on the mobile terminal and starts the on-site navigation measurement mode. All control cross-sections are automatically sorted in upstream and downstream order of the river channel. The mobile terminal sends navigation commands to the drone, which automatically flies to a low altitude of 5 meters above the first control cross-section and hovers there. Simultaneously, the mobile terminal screen displays the current control cross-section number, cross-section center coordinates, and left / right bank positioning markers. It also uses GPS to correct the operator's ground position in real time, guiding the operator to the starting point on the left bank of the control cross-section. After measuring a single control cross-section, the drone automatically flies to the next control cross-section, completing the location guidance for all control cross-sections in a loop.

[0045] Using the current control section as a benchmark, transverse measuring points are evenly distributed according to the pre-set spacing of the measuring points in the section layout list. Starting from the left bank of the section, measuring points are divided equidistantly along the river channel, covering the left bank, left 1 / 6, left 1 / 3, midpoint, right 1 / 3, right 1 / 6, and right bank. For complex sections such as bends, tributaries, and bridges, the number of measuring points is increased to 7, while 5-6 measuring points are set for straight river sections. The mobile terminal displays the horizontal distance of each measuring point from the left bank in real time. The operator sequentially reaches each measuring point along the transverse section to measure siltation, measuring the same point three times and removing outliers. The average of the three sets of data is calculated as the final siltation thickness at that point. The measurement data is automatically bound to the measuring point number, section number, measured verticality, and measurement time, and stored locally, eliminating the need for manual recording.

[0046] After all measuring points of a control section have been measured, the mobile terminal automatically generates a measuring point data table for that control section, including distance from the left bank, three measured depths, average siltation thickness, and verticality compliance indicators. Simultaneously, the standard deviation of the measuring point data for that control section is calculated. If the standard deviation exceeds a threshold, it indicates data dispersion and large measurement errors. A pop-up window prompts the operator to remeasure the abnormal measuring points. Once all measuring points are found to be normal and the verification is passed, the entire set of measurement data for that control section is locked and stored in the section dataset. The measurement of all control sections in the section layout list is completed sequentially. The mobile terminal then aggregates all the raw data from all measuring points for all control sections, integrating the section number, river width, total number of measuring points, siltation thickness at each measuring point, average siltation thickness, maximum siltation thickness, and siltation morphology characteristics of the section, outputting structured section measurement data.

[0047] In some embodiments, a portable thickness measuring device is used to measure the silt thickness at each control section. The portable thickness measuring device integrates a damping module, an tilt sensing module, a laser ranging module, and a Bluetooth module. The damping module is used to control the penetration resistance, allowing the portable thickness measuring device to penetrate the silt at a uniform speed. The tilt sensing module is used to monitor the verticality of the rod in real time, allowing the portable thickness measuring device to penetrate the silt vertically. The laser ranging module is used to read the measurement data and upload it to a mobile terminal in real time via the Bluetooth module.

[0048] Please see Figure 2The diagram shows the overall structure of the portable thickness measuring device. The device has a vertical, coaxial columnar structure, with functional components assembled coaxially from top to bottom. It is suitable for on-site measurement of silt thickness in shallow rural river channels. Specifically, a portable damping penetration handle 1 with a T-shaped grip is located at the top of the device. The handle 1 has anti-slip textured surfaces on both sides for the operator to apply vertical downward force. The lower end of the handle 1 is coaxial and connected to a rotary damping adjuster 2 via a threaded structure, transmitting the downward load. The rotary damping adjuster 2 uses a threaded rotary adjustment structure, which can change the flow gap of the internal damping medium by rotation, thereby adjusting the downward penetration resistance of the measuring rod 3, achieving controllable penetration speed. By limiting the downward pressure rate of the silt layer, measurement errors caused by uneven force application are eliminated. The lower end of the rotary damping adjuster 2 is connected to the measuring rod 3, which is a smooth metal straight rod. The lower end of the measuring rod 3 extends into the depth scale 4 and slides downward synchronously with the portable damping penetration handle 1. A depth scale 4 is fitted onto the outside of the measuring rod 3. The rod is printed with uniform depth markings for intuitive reading of the silt penetration depth. A limiting ring is located at the top of the depth scale 4 to limit the maximum downward stroke and prevent the rod from detaching. A conical drill bit 6 is fixed to the bottom of the measuring rod 3. This conical bit has a smooth surface, facilitating the penetration of aquatic plants and soft silt, and can accurately penetrate the silt layer to contact the underlying hard riverbed base, serving as the bottom reference point for measuring silt thickness. An external fixed scale rod 5 is the outer fixing sleeve of the device, enclosing the depth scale 4. The lower end of the external fixed scale rod 5 is fixedly connected to the outer ring of the conical drill bit 6, keeping the entire device stationary. During measurement, only the measuring rod 3 slides downwards. The vertical silt thickness is directly read by the relative displacement between the external fixed scale rod 5 and the scale of the measuring rod 3.

[0049] An inclination sensing module is installed inside the cavity of the portable damping penetration handle 1. The inclination sensing module is coaxially fixed with the measuring rod 3. The inclination sensing module monitors the pitch and roll angles of the measuring rod 3 in real time to ensure that the measuring rod 3 remains vertical. The inclination sensing module can be an inclination sensor. A laser ranging module is integrated and installed at the upper limit stop ring of the depth scale 4. The transmitting end is vertically downward and aligned with the top surface of the tapered drill bit 6 at the bottom. At the same time, a reflective reference surface is set on the upper end of the external fixed scale rod 5. By measuring the relative travel of the moving measuring rod 3 and the external fixed scale rod 5, the accumulation thickness is automatically calculated. The laser ranging module can be a laser ranging sensor. A Bluetooth module is built into the sealed cavity on the side wall of the rotary damping adjuster 2. A small lithium battery is installed synchronously, and a waterproof sealing interface is reserved for transmitting inclination and distance thickness data wirelessly to the mobile device in real time.

[0050] The device operates as follows: The operator holds the device vertically on the surface of the silt at the control section measurement point. After powering on, the laser ranging module automatically zeros, recording the initial distance to the current reflective reference. The tilt sensing module automatically calibrates the vertical zero point, with a tilt angle of 0° representing the standard vertical state. The Bluetooth module automatically pairs with the on-site mobile terminal. The internal damping resistance is adjusted by rotating the rotary damping adjuster 2, limiting the maximum penetration speed of the measuring rod 3 to ≤5cm / s to avoid measurement distortion caused by excessive manual force. The operator then holds the portable damping penetration handle 1 and applies vertical downward pressure. The measuring rod 3 drives the bottom conical drill bit 6 to penetrate the silt layer at a uniform speed. During this process, the tilt sensing module continuously collects the tilt angle of the rod in real time. If the tilt deviation is >±3°, the Bluetooth module immediately sends an alarm signal to the mobile terminal, which prompts the operator to straighten the device via a vibration pop-up. Simultaneously, the laser ranging module continuously collects the relative displacement value of the moving measuring rod 3 as it slides downward, generating real-time silt thickness data. When the conical drill bit 6 penetrates the silt and contacts the hard riverbed, the downward resistance increases sharply, and the operator stops applying force. At this point, the laser ranging module locks the final displacement value as the single measured silt thickness at that measuring point. The tilt angle sensing module simultaneously records the average tilt angle throughout the entire process for data quality assessment. The Bluetooth module transmits a complete set of data—measuring point number, measured silt thickness, maximum tilt angle, and measurement timestamp—to the mobile terminal in real time. The mobile terminal assesses the data quality; if the tilt angle exceeds a threshold, the data set is marked as suspicious, prompting the operator to remeasure. If the tilt angle is acceptable, the data for that measuring point is automatically saved. The same measuring point is penetrated and measured three times. The Bluetooth module uploads three sets of silt thickness data each time. The mobile terminal automatically removes outliers and calculates the average of the three sets of data as the final silt thickness at that measuring point, automatically incorporating it into the current control section measurement dataset.

[0051] S5. Based on the cross-sectional measurement data, the siltation thickness of each control cross-section is adaptively fitted using an adaptive optimization model, and the siltation volume of the river channel is obtained by combining integral calculations.

[0052] In some embodiments, based on the cross-sectional measurement data, the sedimentation thickness of each control cross-section is adaptively fitted using an adaptive optimization model, and the sedimentation amount of the river channel is obtained by combining integral calculations, including the following steps: S51, the cross-sectional measurement data of each control section are fitted using a weighted parabolic model to generate the cross-sectional sedimentation thickness curve of each control section. S52, perform lateral numerical integration on the cross-sectional sedimentation thickness curve to obtain the cross-sectional sedimentation area corresponding to each control section; S53, with the longitudinal mileage of each control section as the independent variable, constructs independent non-uniform cubic spline interpolation functions for the model parameters, section width and cross-sectional siltation area respectively, and generates the cross-sectional siltation area curve at any longitudinal position of the river channel. S54. Based on the cross-sectional sedimentation area curve, the total sedimentation volume is calculated using the adaptive composite Simpson integral.

[0053] Individual calculations are performed for each control section to fit the complete cross-sectional sedimentation thickness curve. In this embodiment, the adaptive optimization model uses a weighted parabolic model (quadratic). This weighted parabolic model is determined based on the scouring and sedimentation patterns of rural rivers. When local disturbances are ignored, the cross-sectional sedimentation thickness usually follows a distribution that is thicker in the middle and thinner at both banks, which is highly consistent with the parabolic shape. Compared with higher-order polynomials, the quadratic parabolic model has fewer parameters and stronger noise resistance and robustness for sparse measurement points (only 5-7 points per section), which can effectively prevent overfitting.

[0054] Specifically, the weighted parabolic model uses the fundamental function of a quadratic parabola, f(x) = ax² + bx + c, where x represents the lateral distance of the measuring point from the left bank of the river, and f(x) represents the siltation thickness at the corresponding location of the measuring point. This model matches the natural erosion and deposition morphology of rural rivers, where the silt is thicker in the middle and thinner on both banks. The banks are significantly disturbed by water flow and vegetation, making them key control points for riverbed morphology. Therefore, higher weights are assigned to measuring points closer to the left and right banks to enhance the constraint of the curve on the riverbanks and avoid curve distortion caused by relying solely on data from the midpoint of the river. By substituting all the measuring point data of the control section, the optimal coefficients a, b, and c are solved to generate a continuous cross-sectional siltation thickness curve for the entire control section; simultaneously, the goodness of fit R is automatically calculated. 2 If the goodness of fit R 2 If the measurement point is below the set lower limit, it indicates that there are insufficient measurement points for the control section, and additional measurements can be taken on-site. Then, the cross-sectional sedimentation thickness curve of the fitted section is integrated laterally to obtain the cross-sectional sedimentation area S of the control section.

[0055] Using all control sections as longitudinal control nodes, the cross-sectional sedimentation morphology at any location along the entire river channel is reconstructed. This involves extracting parameters from each control section, including the distance *s* along the river, parabolic parameters (a, b, c), cross-sectional width *B*, and cross-sectional sedimentation area *S*. Using the distance *s* as the independent variable, independent non-uniform cubic spline interpolation functions *P(s)* are constructed for each of the five sets of parameters (a, b, c, B, S) to generate the cross-sectional sedimentation area at any longitudinal location along the river channel. Since the river channel has only a few discrete control sections, the complete cross-sectional sedimentation area curve at any longitudinal location along the river channel can be calculated through interpolation, thus expanding the discrete sections into a continuous sedimentation surface for the entire river channel.

[0056] In some embodiments, an adaptive encryption mechanism is introduced when calculating the total sedimentation volume. The convergence of the results is judged by integrating the river channel in segments and calculating the relative error between the total sedimentation volume before and after encryption. The converged value is then rounded to obtain the total sedimentation volume of the entire river channel.

[0057] An adaptive composite Simpson integral is used to calculate the total sedimentation volume. Piecewise integration and adaptive densification convergence judgment automatically balance calculation accuracy and speed. Specifically, the initial number of longitudinal segments is set according to the total river length, dividing the entire river into several small segments along the longitudinal mileage. An interpolation function is used to obtain the sedimentation area of ​​the cross-section at the midpoint of each segment. For each small segment, the sedimentation areas at both ends of the segment are integrated, and the sum is used to obtain the initial total sedimentation volume V0. The number of longitudinal segments is doubled, and the densified total sedimentation volume V1 is recalculated. The relative error between the two volume results is calculated. If the relative error is less than 1%, the calculation result is considered converged, and the iteration stops. If the relative error exceeds the standard, the segment densification is continuously doubled until the accuracy requirement is met. The converged volume value is rounded to obtain the total sedimentation volume of the entire river. The mobile terminal synchronously outputs the results, including the cross-sectional sedimentation area, the sedimentation volume of each river segment, the average sedimentation thickness of the cross-section, the maximum sedimentation thickness, and a goodness-of-fit statistics table, supporting on-site viewing and export / saving.

[0058] Please see Figure 3 As shown in the illustration, another aspect of this application provides a low-cost and rapid system for measuring the siltation volume of rural waterways, the system comprising: The image acquisition module is used to acquire orthophotos of the river channel by flying along the center of the river channel with a drone, and to preprocess the orthophotos to obtain a preprocessed image. The cross-section feature recognition module is used to segment the water area of ​​the preprocessed image to identify the water body range and extract shoreline feature points within the water body range to determine the river centerline. The cross-section screening module is used to screen control cross-sections based on the river centerline to obtain a cross-section layout list; The cross-section measurement module is used to measure the siltation thickness of each control section according to the cross-section layout list and the cross-section location coordinates, so as to obtain the cross-section measurement data of each control section. The siltation calculation module is used to adaptively fit the siltation thickness of each control section based on the cross-sectional measurement data and an adaptive optimization model, and then combine the integral calculation to obtain the siltation volume of the river channel.

[0059] The following provides a specific embodiment to illustrate the solution of this application in detail.

[0060] A rural river is 2.85 km long and 8.5 m to 24.6 m wide, with an average width of about 16.2 m. The river has three bends, including one sharp bend with a turning radius of about 18 m. There are two agricultural bridges and one tributary confluence. The riverbed is visibly raised due to siltation, with dense aquatic plants in some sections and exposed silt in others.

[0061] First, a field survey of the rural river channel was conducted to determine the takeoff and landing points for the drone. In this embodiment, the operator selected an open area at the downstream outlet of the rural river channel as the takeoff and landing points for the drone, as this location offers a wide view and is free of high-voltage power lines. Simultaneously, the drone's battery level and memory card were checked, and the mobile terminal was connected to the GPS beacon. Then, the drone's flight path was planned and controlled via the mobile terminal app for aerial photography. Specifically, the drone was controlled to fly along the center of the river channel at a height of 50 meters, ensuring an image resolution better than 5 cm / pixel. The flight path was approximately 3.2 km long (including extensions at both ends of the river channel). Through automatic flight, a total of 156 photos were acquired, which constitute the orthophotos of the river channel. The photos taken by the drone can be viewed in the newly created "[River Name]" project on the mobile terminal app.

[0062] Then, the mobile app automatically identifies the water body range by segmenting the orthophoto image; it extracts the left and right bank boundary lines through edge detection, calculates the line connecting the midpoints of the left and right boundaries, and generates a center line after cubic spline smoothing. The operator verifies whether the obtained center line matches the actual river channel direction and makes local fine-tuning, for example, in this embodiment, two local fine-tuning points are made at the tributary confluence. After fine-tuning, the mobile app automatically selects control sections by calling the filtering rules, where the parameters are set as follows: fixed spacing ΔL = 500m, automatically recommended by the system based on the river length of 2.85km; curvature threshold is 0.05m. -¹ The width change rate threshold is 0.3. The results obtained after screening include: ① 6 equidistant cross-sections with distances along the river: s=0m (starting point), s=500m, s=1000m, s=1500m, s=2000m, and s=2500m; ② 1 abrupt change in morphology, where the river width suddenly narrows from 18.5m to 11.2m at s=780m; ③ 2 bends, one with a curvature of 0.062 at s=320m and the other with a curvature of 0.058 at s=1650m; ④ 2 bridges added for water-related structures, located at s=210m and s=2630m respectively. The final total number of candidate cross-sections is 6+1+2+2=11. After operator confirmation, the mobile terminal App generates a list of control cross-sections for a certain river, as shown in Table 1 below.

[0063] Table 1

[0064] Next, the operator clicks "Start Measurement" on the mobile terminal app, automatically guiding the drone to each cross-section. After takeoff, the drone automatically flies to a height of 5m above the DM-01 cross-section and hovers. At this time, the mobile terminal displays "Current Location: DM-01 Cross-section, Left Bank Coordinates (Latitude and Longitude)". The measurement process at other cross-sections is illustrated in this embodiment using the DM-03 bend cross-section as an example. At the DM-03 bend cross-section, the position is s=320m (bend apex), the river width is 18.1m, and there are 7 preset measurement points, including the left bank, left 1 / 6, left 1 / 3, midpoint, right 1 / 3, right 1 / 6, and right bank. The specific data during the measurement process are shown in Table 2 below.

[0065] Table 2

[0066] During the measurement process, data quality was fed back in real time. For example, when measuring point 3 was first penetrated, the verticality was 3.5°, exceeding the threshold. At this time, the mobile terminal APP vibrated and prompted "Verticality deviation is too large, it is recommended to remeasure". After the operator adjusted, the remeasurement was qualified. All cross-sectional measurement data are shown in Table 3 below, totaling 11 cross-sections and 68 measuring points.

[0067] Table 3

[0068] Finally, data from 68 measuring points were uploaded to a mobile terminal via Bluetooth. The operator clicked "Calculate Total Silt Content," and the mobile terminal app invoked the adaptive optimization model for calculation, with the calculation progress visible. First, weighted parabolic fitting was performed on each of the 11 cross-sections, with a goodness-of-fit R² of 0.86 (e.g., DM-05), a maximum of 0.97 (e.g., DM-08), and an average of 0.92. Then, longitudinal interpolation using non-uniform cubic splines was performed, with interpolation nodes being the distance *s* along the 11 cross-sections and the corresponding model parameters (a, b, c, B, A), and 100 longitudinal discrete points M (automatically set, step size approximately 28.5 m). Finally, the total silt content was calculated using volume integration (e.g., the compound Simpson's rule). The initial step size Δs = 28.5m was set, the integral value V0 = 13,285.6m³, and M was doubled to 200 to calculate the integral value V1 = 13,316.2m³. The relative change was 0.23% < 1%, which indicates convergence. The final total sedimentation volume was 13,305m³ (accurate to 1m³).

[0069] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0070] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0071] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0072] Please see Figure 4 , Figure 4 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 401 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 402 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 402 can store the operating system and other applications. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 402 and is called and executed by the processor 401 using the methods described in the embodiments of this application. Input / output interface 403 is used to implement information input and output; The communication interface 404 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 405 transmits information between various components of the device (e.g., processor 401, memory 402, input / output interface 403, and communication interface 404); The processor 401, memory 402, input / output interface 403 and communication interface 404 are connected to each other within the device via bus 405.

[0073] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0074] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0075] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0076] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0077] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0078] This application provides a low-cost and rapid method and system for measuring the siltation of rural waterways. The scheme acquires orthophotos of the waterway via UAV flight and performs preprocessing. Compared to full-area grid aerial photography, this significantly reduces the number of images collected and shortens the time spent on field aerial photography. It is suitable for narrow and small rural waterways. Furthermore, preprocessing eliminates interference from water surface reflections, tree shadow distortion, segmented color aberration, and lens distortion, outputting standardized images with a unified geographic coordinate system. Then, a lightweight semantic segmentation network is used to achieve pixel-level water body recognition, combined with morphological optimization, accurately distinguishing the main waterway, scattered water accumulation on the banks, and floating duckweed and shallows. The river's water boundaries are fully locked, avoiding interference from non-target water areas. By extracting the inflection points and abrupt changes in width of the shorelines, and fitting the midpoints of paired shorelines to generate a smooth and continuous centerline, the true course of the river is accurately reconstructed. Then, control sections are evenly distributed in straight river sections. Control sections are automatically added in areas prone to siltation, such as sharp bends, abrupt changes in river width, tributary confluences, and bridge / dam areas. The number of measuring points is allocated according to the section type to balance survey workload and measurement accuracy, outputting a structured section layout list. This list is then directly distributed to drones for measuring point navigation, solving the problems of traditional manual point placement lacking precise coordinates and difficulty in locating sections in the field. Drone navigation locates the control section measuring points, and a lightweight thickness measuring device is used to measure siltation thickness. The measurement data is reproducible and highly reliable. An adaptive optimization model is used to fit the measurement data, and integral calculations are combined to obtain the river's siltation volume. From aerial map generation, automatic point placement, high-precision field measurement to automatic calculation of siltation volume across the entire area, the entire process significantly reduces the workload of manual surveying, mapping, recording, and indoor calculation. The entire solution can be completed using only low-cost drones, portable measuring rods, and ordinary mobile terminals. It is suitable for rural shallow water, aquatic plants, and narrow river conditions, and its measurement accuracy is superior to traditional manual cross-section survey methods, greatly shortening the river siltation survey cycle.

[0079] Furthermore, this application also has the following advantages: Based on UAV orthophotos, this application automatically identifies the river channel centerline and intelligently selects control sections. Section selection is automatically completed through image processing, eliminating reliance on manual field line selection and significantly reducing preliminary fieldwork. The lightweight thickness measurement device integrates damping penetration control and tilt angle sensing calibration design. Through an adjustable damping mechanism and tilt angle sensor, it solves the problem of inconsistent force and verticality during manual pole insertion, exhibiting strong environmental adaptability and high data reliability. Based on a longitudinal and transverse adaptive optimization algorithm using a small number of control sections, it utilizes the regularity of river channel sedimentation spatial distribution to invert the entire river channel sedimentation surface through sparse section data, achieving quantitative estimation from point data to three-dimensional total sedimentation. The application integrates a closed-loop design on a mobile terminal, combining UAV image recognition, section selection, measurement point guidance, data recording, and volume calculation. This integrates distributed functions into a single-handed mobile terminal, enabling immediate results from on-site measurements. Based on the parabolic siltation distribution hypothesis and weighted fitting method for rural river characteristics, this paper fully utilizes prior laws and combines a weighted strategy of near-shore measuring points to improve the fitting accuracy of cross-sectional siltation curves.

[0080] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0081] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0082] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0083] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application 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 the embodiments of this application 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.

[0084] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A low-cost and rapid method for measuring the siltation volume of rural waterways, characterized in that, The method includes: The orthophoto of the river is obtained by flying along the center of the river with a drone, and the orthophoto is preprocessed to obtain a preprocessed image. The preprocessed image is segmented to identify the water body range, and shoreline feature points within the water body range are extracted to determine the river centerline; Based on the centerline of the river channel, control sections are selected to obtain a list of section layouts; Based on the cross-section layout list, the siltation thickness of each control cross-section is measured in conjunction with the cross-section location coordinates to obtain the cross-section measurement data of each control cross-section. Based on the cross-sectional measurement data, the sedimentation thickness of each control cross-section is adaptively fitted using an adaptive optimization model, and the sedimentation amount of the river channel is obtained by combining integral calculations.

2. The method according to claim 1, characterized in that, The preprocessed image is segmented to identify the water body area, and shoreline feature points within the water body area are extracted to determine the river centerline, including: Extract multi-channel features from the preprocessed image to construct a water body feature discrimination dataset; Based on the water body feature discrimination dataset, the preprocessed image is classified pixel-by-pixel using a lightweight semantic segmentation network, and morphological constraints are added to obtain a water body mask. Extract the closed outer contour of the water body mask to obtain the water body range of the entire river channel; Extract feature points from both sides of the shoreline within the water body to obtain a set of feature points on both banks; Along the direction of the river flow, the feature points in the feature point set on both banks are paired one by one according to the transverse direction of the river, forming multiple sets of transverse lines that cross the river. Calculate the spatial midpoint coordinates of the two feature points of the transverse line, and fit the spatial midpoint coordinates based on cubic spline interpolation to obtain the river centerline.

3. The method according to claim 2, characterized in that, The lightweight semantic segmentation network is a lightweight encoder-decoder symmetric semantic segmentation network. The encoder backbone adopts an improved MobileNetV2 lightweight inverse residual structure as a feature extractor, and the decoder adopts a skip connection, channel concatenation, and lightweight fusion convolution structure to restore the image resolution layer by layer. The output pixel segmentation probability map is activated by the Sigmoid function.

4. The method according to claim 1, characterized in that, Based on the selection of control sections according to the river centerline, a list of section layouts is obtained, including: Using the starting point of the river centerline as the origin of the coordinate system, cross sections perpendicular to the direction of the river are generated at fixed intervals along the longitudinal direction of the water flow on the river centerline, serving as basic equidistant candidate cross sections. The changes in river morphology are determined segment by segment along the centerline of the river channel. If any judgment condition is met, a new control section is added. The basic equidistant candidate sections and newly added control sections are deduplicated and verified, and the number of measuring points is allocated to generate a section layout list.

5. The method according to claim 4, characterized in that, The determination criteria include the determination of bend curvature, the determination of sudden changes in river width, the determination of tributary confluence, and the determination of water-related structures.

6. The method according to claim 1, characterized in that, The thickness of silt deposits at each control section is measured using a portable thickness measuring device. This device integrates a damping module, an inclination sensing module, a laser ranging module, and a Bluetooth module. The damping module controls the penetration resistance, ensuring the device penetrates the silt at a uniform speed. The inclination sensing module monitors the verticality of the rod in real time, ensuring the device penetrates the silt vertically. The laser ranging module reads the measurement data and uploads it to a mobile terminal in real time via Bluetooth.

7. The method according to claim 1, characterized in that, Based on the cross-sectional measurement data, the sedimentation thickness of each control cross-section is adaptively fitted using an adaptive optimization model, and the sedimentation amount of the river channel is obtained by combining integral calculations, including: The cross-sectional measurement data of each control section were fitted using a weighted parabolic model to generate the cross-sectional sedimentation thickness curves of each control section. The cross-sectional sedimentation thickness curve is subjected to lateral numerical integration to obtain the cross-sectional sedimentation area corresponding to each control section. Using the longitudinal mileage of each control section as the independent variable, independent non-uniform cubic spline interpolation functions are constructed for the model parameters, section width, and cross-sectional siltation area, respectively, to generate the cross-sectional siltation area curve at any longitudinal position of the river channel. The total amount of sedimentation is calculated using the adaptive composite Simpson integral based on the cross-sectional sedimentation area curve.

8. The method according to claim 7, characterized in that, An adaptive densification mechanism is introduced when calculating the total sedimentation volume. The convergence of the result is judged by integrating the river channel in segments and calculating the relative error between the total sedimentation volume before and after densification. The converged value is rounded to obtain the total sedimentation volume of the entire river channel.

9. A low-cost, rapid system for measuring siltation in rural waterways, applied to the method described in any one of claims 1-8, characterized in that, The system includes: The image acquisition module is used to acquire orthophotos of the river channel by flying along the center of the river channel with a drone, and to preprocess the orthophotos to obtain a preprocessed image. The cross-section feature recognition module is used to segment the water area of ​​the preprocessed image to identify the water body range and extract shoreline feature points within the water body range to determine the river centerline. The cross-section screening module is used to screen control cross-sections based on the river centerline to obtain a cross-section layout list; The cross-section measurement module is used to measure the siltation thickness of each control section according to the cross-section layout list and the cross-section location coordinates, so as to obtain the cross-section measurement data of each control section. The siltation calculation module is used to adaptively fit the siltation thickness of each control section based on the cross-sectional measurement data and an adaptive optimization model, and then combine the integral calculation to obtain the siltation volume of the river channel.

10. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method according to any one of claims 1 to 8.

Images