Water and soil loss remote sensing dynamic monitoring method based on unmanned aerial vehicle

Abstract

The invention relates to the field of image monitoring, in particular to a water and soil loss remote sensing dynamic monitoring method based on an unmanned aerial vehicle, and the method comprises the steps: collecting an aerial image, extracting basic landform visual features in the aerial image, recognizing a significant boundary, constructing a reference window based on the significant boundary, and locking a variation local reference window. And verifying the variation local reference window based on the sample features to judge whether to mark the aerial image, subsequently identifying the image by using the waste slag segmentation model, analyzing the industrial waste slag interception rate, and subsequently determining a corresponding intervention measure. According to the method, by considering the visual features of the local landform, the abnormal local reference window is quickly locked, a large number of aerial images without effective information are removed, the aerial images with high tendency of the abandoned slag area are focused, and then the aerial images are accessed to the abandoned slag segmentation model for analysis; therefore, the remote sensing monitoring reliability is ensured while the data processing amount during the processing of massive aerial images is reduced, and the water and soil loss monitoring of a large-area monitoring area is realized.

Description

Technical Field

[0001] This invention relates to the field of image monitoring, and more particularly to a method for dynamic remote sensing monitoring of soil erosion based on unmanned aerial vehicles (UAVs). Background Technology

[0002] In monitoring soil and water loss in mining areas, the waste interception rate is crucial. It measures the effectiveness of engineering measures (such as waste dams, sedimentation basins, and soil covering and revegetation) in controlling loose waste, which is the main source of soil and water loss. A higher interception rate indicates less loose material exposed and easily washed away by water flow, thus lowering the risk of soil and water loss. Traditional methods for monitoring soil erosion mainly rely on ground observations and waste disposal model simulations, which are subject to subjectivity and errors, and cannot acquire dynamic change data in real time, making it difficult to reflect the dynamic process of soil erosion in a timely manner. With the continuous development of UAV technology, its application in soil erosion monitoring is also constantly expanding and deepening. Combining UAV remote sensing data with satellite remote sensing data can achieve more comprehensive and accurate monitoring. By using machine learning and artificial intelligence algorithms, the data collected by UAVs can be automatically analyzed and processed to improve monitoring accuracy and efficiency. Regularly acquiring image data of the monitoring area allows for observation of soil erosion trends, identification of erosion-sensitive areas, and provides a basis for the formulation and optimization of control measures.

[0003] Chinese Patent Application Publication No. CN114113541A discloses a soil erosion monitoring device, belonging to the field of soil erosion technology. It includes a first baffle and side baffle mechanisms. Side baffle mechanisms are fixedly connected to both the upper and lower sides of the right side of the first baffle. A soil collection mechanism is arranged between the two side baffle mechanisms and is fixedly connected to the first baffle. An adjustment mechanism is provided on the first baffle, located within a cavity opened inside the first baffle. In this invention, by setting up a first baffle and side baffle mechanisms, the first baffle, in conjunction with a second baffle and a U-shaped shell, can effectively block eroded soil, achieving the purpose of collecting eroded soil and facilitating subsequent calculations of the degree of soil erosion on highway slopes. Simultaneously, the U-shaped plate is no longer fixed, allowing the U-shaped shell to slide freely outside the second baffle. When the U-shaped rod is released, the spring can extend using its own elasticity, thereby pushing the support plate to tightly lock the locking block into the slot, fixing the U-shaped shell.

[0004] Chinese Patent Application Publication No. CN119064554A discloses a method for monitoring and predicting soil erosion in small watersheds. This method can predict soil erosion through an established mathematically universal waste disposal model, providing data support for the implementation of soil erosion control. A collection device and a data acquisition device are set up in the selected monitoring site. The collection device collects slope runoff to a single outlet, while the data acquisition device collects rainfall and suspended mass separately. Manual weighing yields bedload data. By collecting rainfall, suspended mass, and bedload over a period of time, multiple sets of erosion data can be obtained. Correlation analysis is used to fit curves to these multiple sets of erosion data, resulting in a mathematically universal waste disposal model. This model can predict soil loss during rainfall periods based on the rainfall and suspended mass collected by the data acquisition device, facilitating the development of corresponding soil erosion control measures according to different situations.

[0005] However, the above method has the following problems: when faced with a large number of aerial images, due to terrain factors, there may be a large number of similar aerial images or aerial images that do not contain effective information. The method of using visual model to traverse and analyze is extremely computationally expensive and has low efficiency in analyzing waste disposal areas. Summary of the Invention

[0006] To address this, the present invention provides a method for dynamic monitoring of soil erosion using remote sensing based on unmanned aerial vehicles (UAVs), which overcomes the problems of sparse distribution of ground monitoring stations in existing technologies, making it difficult to fully cover large areas, acquire dynamic change data in real time, and reflect the dynamic process of soil erosion in a timely manner.

[0007] To achieve the above objectives, the present invention provides a method for dynamic monitoring of soil erosion using remote sensing based on unmanned aerial vehicles (UAVs), comprising: The drone is used to collect aerial images of the monitoring area according to a preset flight path; Basic landform visual features are extracted from aerial images acquired within a predetermined time domain; Identify salient boundaries in aerial images, construct reference windows based on salient boundaries, determine local reference windows segmented by salient boundaries, and enhance local images within local reference windows; By comparing the enhanced local image with the basic terrain visual features, the aberrant local reference window is locked, and the aberrant local reference window is verified based on the sample features to determine whether the aerial image should be marked. The analysis of the marked aerial images includes using a waste disposal segmentation model to identify and segment the waste disposal area and the utilized waste disposal area in the aerial images in order to determine the corresponding waste disposal area and utilization area. Calculate the industrial waste interception rate in the monitoring area, and determine whether to set up a sedimentation tank in the monitoring area based on the comparison result between the industrial waste interception rate and the interception rate threshold.

[0008] Furthermore, the process of extracting basic landform visual features based on aerial images acquired within a predetermined time domain includes, Extract the chromaticity values ​​and texture density from each frame of aerial images within a predetermined time domain; The mean chromaticity value and the mean texture density are calculated as the basic visual features of the terrain.

[0009] Furthermore, the process of identifying salient boundaries in aerial images, constructing reference windows based on these salient boundaries, and determining the local reference windows segmented by these salient boundaries includes: The contour lines in the aerial image are determined. If the difference in chromaticity values ​​on both sides of the contour line is greater than a predetermined chromaticity difference threshold, the contour line is determined as a significant boundary. A reference window is overlaid on a significant boundary such that the significant boundary passes through the reference window, thereby dividing the reference window into local reference windows.

[0010] Furthermore, the process of locking the aberrant local reference window by comparing the enhanced local image with the basic topographical visual features includes: Extract the visual features of the terrain within the local reference viewports; Compare the visual features of landforms with the visual features of basic landforms; If the condition of topographic difference is met, the local reference window is determined to be an aberrant local reference window; Among them, the landform difference condition is that the chromaticity difference ratio is greater than a predetermined chromaticity difference ratio threshold and the texture density difference ratio is greater than a predetermined texture density difference ratio threshold.

[0011] Furthermore, the process of determining whether to label aerial images includes, Obtain sample features, including sample chromaticity values ​​and sample texture density; The mutation reference window is validated, including extracting the chromaticity value and texture density value within the mutation reference window; Determine the chromaticity difference ratio of the chromaticity value relative to the sample chromaticity value, and determine the texture density difference ratio of the texture density value relative to the sample texture density value; If the chromaticity difference ratio is less than a preset sample feature difference ratio threshold and the texture density difference ratio is less than a preset sample texture feature difference ratio threshold, then it is determined that the aerial image containing the variant reference window should be marked.

[0012] Furthermore, the steps by which the waste disposal model learns from the image pre-data include: The learning parameters of the waste disposal model were adjusted to standard parameters; The image pre-data is then passed into the waste disposal segmentation model; The waste disposal segmentation model learns from the image pre-data, identifies the waste disposal area and the utilized waste disposal area, and marks them.

[0013] Furthermore, the steps for determining the waste disposal area and the utilization area include: The boundary contours of the waste disposal area and the utilized waste disposal area are extracted using an edge detection algorithm. The boundary contour is rendered and fitted to obtain the corresponding closed curve; The area enclosed by the closed curve is calculated using an integral method to obtain the corresponding waste area and utilization area.

[0014] Furthermore, the industrial waste interception rate of the monitoring area is calculated based on the waste disposal area and the utilization area, wherein, The industrial waste interception rate is the quotient of the waste area and the utilization area.

[0015] Furthermore, the industrial waste interception rate is compared with the interception rate threshold. When the industrial waste interception rate is greater than the interception rate threshold, the monitoring area is not adjusted.

[0016] Furthermore, when the industrial waste interception rate is less than the interception rate threshold, a sedimentation tank is installed in the monitoring area, wherein... The sedimentation tank is used to guide industrial waste from the waste disposal area to the utilized waste disposal area.

[0017] Compared with existing technologies, this invention acquires aerial images, extracts basic topographical visual features within the images, identifies significant boundaries, constructs reference windows based on these boundaries, locks onto anomalous local reference windows, and verifies these anomalous local reference windows based on sample features to determine whether the aerial images should be marked. Subsequently, a waste disposal segmentation model is used to identify the images, analyze the industrial waste disposal interception rate, and determine corresponding intervention measures. This invention, by considering local topographical visual features, quickly locks onto anomalous local reference windows, eliminates a large number of aerial images containing no effective information, focuses on aerial images with a high tendency to contain waste disposal areas, and then integrates them into a waste disposal segmentation model for analysis. This reduces the data processing volume when processing massive amounts of aerial images while ensuring the reliability of remote sensing monitoring, enabling the monitoring of soil erosion in large monitoring areas.

[0018] In particular, this invention acquires the basic geomorphic visual features of aerial images, identifies significant boundaries, and constructs local reference windows for image enhancement to lock onto anomalous local reference windows. In remote sensing mapping, large areas are typically monitored, resulting in a massive amount of aerial images. Reading these images traversally is extremely computationally expensive. In practice, the geomorphic visual features of continuous areas are often similar, such as plains and deserts. Many aerial images may not contain waste disposal areas. Due to the similarity of geomorphic visual features, aerial images often appear similar or identical. Therefore, this invention considers prior identification of the basic geomorphic visual features of the current area as a foundation, clearly defining significant boundaries in the aerial images. Significant boundaries typically appear when closed regions are present in the image. Only local reference windows near significant boundaries are considered, initially reflecting visual similarities to the geomorphic visual features. In areas exhibiting anomalies, the extraction of basic topographical visual features is relatively fast and computationally efficient, allowing for quick focusing or capture of regions visually different from the basic topographical features. These regions may be waste disposal areas. Therefore, without intervening in the visual model to read aerial images, aerial images containing no effective information are filtered out. Subsequently, a reference window for anomalies is used for verification to further exclude local areas that differ significantly from waste disposal areas. Then, aerial images with a high probabilities of containing waste disposal areas are marked, and then the visual model is used for analysis. When dealing with massive amounts of aerial images, this approach can, while ensuring reliability, utilize basic topographical visual features to remove a large number of aerial images without effective information, and allow the visual model to focus its attention on the marked aerial images, thereby improving the monitoring efficiency of waste disposal areas in aerial images.

[0019] Furthermore, by refining the steps for drones to collect aerial images of the monitoring area, the accuracy, real-time performance, automation, and security of monitoring can be significantly improved. At the same time, resource utilization can be optimized, and the richness and traceability of data can be enhanced, making drone remote sensing technology more advantageous in soil and water conservation monitoring and providing stronger support for environmental protection and resource management.

[0020] Furthermore, the pixel filtering process can effectively improve the quality of aerial images, enhance monitoring accuracy, optimize data processing efficiency, improve the training effect of waste disposal models, and enhance the reliability and stability of the monitoring system. At the same time, it reduces subsequent processing costs, making the entire monitoring system more efficient, accurate, and economical, and better able to meet the needs of dynamic monitoring of soil and water loss.

[0021] Furthermore, by iteratively optimizing, testing, and saving the training parameters of the waste disposal model, the accuracy, reliability, and performance optimization of the model can be ensured. This rigorous training process not only improves the performance of the waste disposal model in practical applications but also enhances the system's scalability, stability, and operational efficiency, while reducing operating costs.

[0022] Furthermore, by adjusting the learning parameters of the waste disposal model, inputting pre-image data, and labeling it, the consistency and accuracy of the waste disposal model can be ensured, the learning efficiency can be improved, the generalization ability of the waste disposal model can be enhanced, the automation level of the system can be increased, the data processing flow can be optimized, and the reliability and stability of the system can be enhanced.

[0023] Furthermore, by using edge detection algorithms to extract boundary contours, rendering and fitting the boundary contours, and using integral methods to calculate the area, the boundaries between the waste disposal area and the utilized waste disposal area can be accurately extracted, improving the accuracy of area calculation, enhancing the automation and flexibility of the system, and reducing system operating costs.

[0024] Furthermore, by calculating the industrial waste interception rate and making decisions based on the comparison results, the effectiveness of waste interception can be accurately assessed, enabling scientific decision-making, effectively controlling soil erosion, and making the entire monitoring system more efficient, accurate, and economical, thus better meeting the needs of dynamic monitoring of soil erosion. Attached Figure Description

[0025] Figure 1 This is a flowchart of the method for dynamic monitoring of soil erosion based on remote sensing using unmanned aerial vehicles (UAVs) according to an embodiment of the present invention. Figure 2 A logic block diagram for determining significant boundaries in an embodiment of the invention; Figure 3 A logic block diagram for determining the variable local reference window in an embodiment of the invention; Figure 4 This is a logic block diagram illustrating how to determine whether to mark aerial images according to an embodiment of the invention. Detailed Implementation

[0026] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0027] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0028] It should be noted that in the description of this invention, the terms "upper," "lower," "left," "right," "inner," and "outer," etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is merely for description and does not indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0029] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0030] Please see Figure 1 As shown, it is a flowchart of the method for dynamic monitoring of soil erosion based on UAV remote sensing according to an embodiment of the present invention, including: Step S1: Use a drone to collect aerial images of the monitoring area according to a preset flight path; Step S2: Extract basic landform visual features based on aerial images acquired within a predetermined time domain; Step S3: Identify salient boundaries in the aerial image, construct reference windows based on the salient boundaries, determine the local reference windows segmented by the salient boundaries, and enhance the local image within the local reference windows; Step S4: Based on the comparison between the enhanced local image and the basic landform visual features, lock the abnormal local reference window, and verify the abnormal local reference window based on the sample features to determine whether to mark the aerial image. Step S5: Analyze the marked aerial images, including using a waste disposal segmentation model to identify and segment the waste disposal area and the utilized waste disposal area in the aerial images, so as to determine the corresponding waste disposal area and utilization area. Step S6: Calculate the industrial waste interception rate of the monitoring area, and determine whether to set up a sedimentation tank in the monitoring area based on the comparison result between the industrial waste interception rate and the interception rate threshold. The sample features were determined based on several sample images containing waste disposal areas.

[0031] Specifically, during implementation, when collecting aerial images, the collected aerial images need to be pixel filtered to retain those with a pixel threshold. The pixel threshold is the minimum standard for the clarity of the aerial image and is inversely correlated with the flight altitude of the drone.

[0032] Specifically, in practice, waste material typically refers to loose solid materials such as waste soil, stone, and slag generated during engineering construction, mining, and other activities. These materials have unstable structures and are easily washed away by rainwater, making them a major source of soil erosion. Waste material dumping areas are areas where waste materials that have not been effectively intercepted or secured are piled up, posing a risk of soil erosion. Traditional waste disposal methods include: constructing waste retaining dams, sedimentation basins, covering with soil and revegetating, and using waste for roadbed filling. The purpose of these measures is to stabilize loose waste, reduce its mobility, and thus control soil erosion. In practice, areas where waste has been utilized refer to areas where industrial waste has been collected, stabilized, treated, or recycled.

[0033] The waste disposal area is the area of ​​the waste disposal zone, and the utilization area is the area of ​​the waste disposal zone that has been utilized.

[0034] Furthermore, the industrial waste interception rate is calculated based on the area of ​​the waste disposal area and the area of ​​the waste disposal area that has been utilized. Its essential purpose is to characterize the proportion of waste that has been disposed of, thereby reflecting the effectiveness of waste interception and disposal.

[0035] Specifically, the process of extracting basic landform visual features based on aerial images acquired within a predetermined time domain includes, Extract the chromaticity values ​​and texture density from each frame of aerial images within a predetermined time domain; The mean chromaticity value and the mean texture density are calculated as the basic visual features of the terrain.

[0036] Specifically, texture density is the ratio of the area corresponding to the texture outline to the total area of ​​the aerial image.

[0037] Specifically, please refer to Figure 2 As shown, Figure 2 This is a logic block diagram illustrating the determination of salient boundaries in an embodiment of the invention. The process of identifying salient boundaries in an aerial image, constructing reference windows based on the salient boundaries, and determining local reference windows segmented by the salient boundaries includes: The contour lines in the aerial image are determined. If the difference in chromaticity values ​​on both sides of the contour line is greater than a predetermined chromaticity difference threshold, the contour line is determined as a significant boundary. A reference window is overlaid on a significant boundary such that the significant boundary passes through the reference window, thereby dividing the reference window into local reference windows.

[0038] Specifically, the purpose of setting a chromaticity difference threshold is to characterize the situation where there is a large difference in chromaticity on both sides of the boundary line. Based on this, the chromaticity difference threshold is determined based on the average chromaticity of the corresponding area on both sides of the boundary line. Usually, to indicate a large difference, the chromaticity difference threshold is set between 0.4 and 0.6 times the average value. In practice, 0.5 times is preferred.

[0039] In implementation, when constructing reference windows, efforts should be made to ensure that the salient boundary is located in the middle of the reference window, and that the areas of the two local reference windows divided by the salient boundary are as equal as possible.

[0040] Specifically, please refer to Figure 3 As shown, Figure 3 This is a logic block diagram illustrating the determination of the aberration local reference window in an embodiment of the invention. The process of locking the aberration local reference window by comparing the enhanced local image with basic terrain visual features includes... Extract the terrain visual features within the local reference view, including chromaticity values ​​and texture density values; Compare the visual features of landforms with the visual features of basic landforms; If the condition of topographic difference is met, the local reference window is determined to be an aberrant local reference window; Among them, the landform difference condition is that the chromaticity difference ratio is greater than a predetermined chromaticity difference ratio threshold and the texture density difference ratio is greater than a predetermined texture density difference ratio threshold.

[0041] In practice, the difference ratio is the ratio of the difference between two values ​​to the mean of the two values.

[0042] In practice, the purpose of setting the chromaticity difference ratio threshold and the texture density difference ratio threshold is to distinguish cases where there is a large difference between the visual features of the local reference window and the basic visual features of the monitored area. Therefore, the chromaticity difference ratio threshold and the texture density difference ratio threshold should not be too small. In practice, the chromaticity difference ratio threshold is selected in the range [0.4, 0.5], preferably 0.4, and the texture density difference ratio threshold is selected in the range [0.5, 0.6], preferably 0.6.

[0043] Specifically, please refer to Figure 4 As shown, Figure 4 The present invention provides a logic block diagram for determining whether to mark an aerial image, and the process of determining whether to mark an aerial image includes the following steps: Obtain sample features, including sample chromaticity values ​​and sample texture density; The mutation reference window is validated, including extracting the chromaticity value and texture density value within the mutation reference window; Determine the chromaticity difference ratio of the chromaticity value relative to the sample chromaticity value, and determine the texture density difference ratio of the texture density value relative to the sample texture density value; If the chromaticity difference ratio is less than a preset sample feature difference ratio threshold and the texture density difference ratio is less than a preset sample texture feature difference ratio threshold, then it is determined that the aerial image containing the variant reference window should be marked.

[0044] Specifically, the sample features are determined based on several sample images containing waste areas. The chromaticity values ​​in several waste areas are extracted, the mean chromaticity value is calculated, and this mean chromaticity value is used as the sample chromaticity value. The texture density values ​​in several waste areas are extracted, the mean texture density value is calculated, and this mean texture density value is used as the sample texture density.

[0045] In practice, the purpose of setting the sample feature difference ratio threshold and the sample texture feature difference ratio threshold is to further verify the situation where the image features in the local reference window of the mutation are similar to the image features of the waste area, so as to eliminate the situation that obviously does not belong to the waste area. In practice, the sample feature difference ratio threshold is selected in the range [0.25, 0.3], preferably 0.25, and the texture density difference ratio is selected in the range [0.3, 0.4], preferably 0.3.

[0046] This invention acquires the basic topographical visual features of aerial images, identifies significant boundaries, and constructs local reference windows for image enhancement to lock onto anomalous local reference windows. In remote sensing mapping, large areas are typically monitored, resulting in a massive amount of aerial images. Reading these images traversally is extremely computationally expensive. In practice, the topographical visual features of continuous areas are often similar, such as plains and deserts. Many aerial images may not contain waste disposal areas. Due to the similarity of topographical visual features, aerial images often appear similar or identical. Therefore, this invention considers identifying the basic topographical visual features of the current area as a foundation, clarifying significant boundaries in the aerial images. Significant boundaries typically appear when closed regions are present in the image. Only local reference windows near significant boundaries are considered, initially reflecting visual similarities to the topographical visual features. In areas of anomaly, the extraction of basic topographical visual features is relatively fast and computationally efficient, allowing for quick focusing or capture of regions that visually differ from the basic topographical features. These regions may be waste disposal areas. Therefore, without intervening in the visual model to read aerial images, aerial images containing no effective information are filtered out. Subsequently, verification is performed on the anomaly local reference window to further exclude local areas that differ significantly from waste disposal areas. Then, some aerial images with a high probability of containing waste disposal areas are marked, and then the visual model is used for analysis. When faced with massive amounts of aerial images, this method can eliminate a large number of aerial images without effective information by utilizing basic topographical visual features while ensuring reliability. It also allows the visual model to focus its attention on the marked aerial images, thereby improving the monitoring efficiency of waste disposal areas in aerial images.

[0047] Specifically, the location information of the monitored area is obtained, and the flight path of the drone is determined. Location information includes the geographic coordinates and topographic information of the monitoring area; When the drone reaches the corresponding flight path, it triggers the image acquisition device to capture aerial images. Aerial images are transmitted to the ground control station in real time and stored via a data link.

[0048] In practice, by acquiring the geographic coordinates and topographic information of the monitoring area, drones can collect images along precise flight paths. This ensures the comprehensiveness and consistency of image acquisition, avoiding the problems of missed areas or duplicate acquisition that may occur in traditional methods. Geographic coordinates and topographic information provide important references for subsequent data analysis. For example, when analyzing soil erosion, combining topographic information can more accurately identify key features such as erosion gullies and slope changes, thereby more precisely assessing the degree and influencing factors of soil erosion.

[0049] Aerial images are transmitted to ground control stations in real time, allowing monitoring personnel to obtain data immediately. This real-time capability is particularly important for dynamic monitoring, such as in extreme weather conditions like torrential rains, enabling timely detection and intervention of sudden soil erosion events. Image storage via data links ensures data integrity and security. The stored data can be used for subsequent detailed analysis and historical comparisons, providing data support for long-term soil erosion monitoring and control.

[0050] Precise flight path planning and automated image acquisition reduce unnecessary data collection, avoid data redundancy, and improve data acquisition efficiency. Meanwhile, real-time transmission and storage capabilities enable rapid data utilization, reduce data processing latency, and further optimize resource utilization.

[0051] Image data collected by combining geographic coordinates and topographic information includes not only visual information but also rich geographic and topographic details. This information supports subsequent multi-dimensional analysis, such as analyzing the influencing factors of soil erosion in conjunction with factors like terrain slope and vegetation cover. All collected data is stored via data links for easy retrieval and traceability. This provides a data foundation for analyzing historical changes in the monitored area, helping to evaluate the effectiveness of control measures and adjust monitoring strategies.

[0052] By further refining the steps for drones to collect aerial images of the monitoring area, the accuracy, real-time performance, automation, and security of monitoring can be significantly improved. At the same time, resource utilization can be optimized, and the richness and traceability of data can be enhanced, making drone remote sensing technology more advantageous in soil and water loss monitoring and providing stronger support for environmental protection and resource management.

[0053] The acquired aerial images need to be pixel-filtered, including: The ground control station acquires pixel values ​​from aerial images; The pixel value is compared with the pixel threshold, where... When the pixel value is greater than the pixel threshold, the aerial image is retained; when the pixel value is less than the pixel threshold, the aerial image is filtered. Collect and preserve aerial images.

[0054] In practice, pixel filtering is used to retain only aerial images with pixel values ​​above the pixel threshold. This means the selected images have higher clarity and resolution, and can more accurately reflect the detailed information of the monitored area, such as the texture, shape, and distribution of the waste disposal area. Images with pixel values ​​below the pixel threshold usually have blurriness, noise, or other quality issues, which may introduce errors in subsequent analysis. Filtering these low-quality images ensures a more reliable data foundation for subsequent processing.

[0055] Aerial images can more clearly reveal the characteristics of waste disposal areas, such as texture density, arrangement regularity, and accumulation height. These features are crucial for subsequent waste disposal area identification and segmentation, significantly improving the recognition accuracy of waste disposal segmentation models. Low-quality images may lead to misclassification or omission of waste disposal areas by the waste disposal segmentation model. Pixel filtering can effectively reduce such problems and improve the accuracy and reliability of monitoring results.

[0056] Filtering out low-quality images significantly reduces the amount of data. This not only saves storage space but also reduces the computational load of subsequent data processing, improving the overall efficiency of the monitoring system. Because of the reduced data volume, steps such as image preprocessing, feature extraction, and waste disposal model training can be performed faster, enabling more timely completion of monitoring tasks and providing quicker support for subsequent decision-making.

[0057] Collected and preserved aerial images form a corresponding aerial image dataset, which is used to train the waste disposal model. High-quality training data can better reflect the true features of the waste disposal area, helping the model learn more accurate patterns and thus improving its generalization ability and prediction performance. Low-quality images may introduce noise and outliers, leading to overfitting of the waste disposal model. Pixel filtering can reduce the impact of such data on the training of the waste disposal model, making it more robust.

[0058] The pixel filtering process can effectively improve the quality of aerial images, enhance monitoring accuracy, optimize data processing efficiency, improve the training effect of waste disposal models, and enhance the reliability and stability of the monitoring system. At the same time, it reduces subsequent processing costs, making the entire monitoring system more efficient, accurate and economical, and better able to meet the needs of dynamic monitoring of soil and water loss.

[0059] Specifically, the steps for preprocessing aerial images include: The aerial images are standardized and corresponding image pre-data is generated; The standardization process involves segmenting the aerial image according to a standard segmentation ratio. The standard segmentation rate is the segmentation rate that the waste disposal model can recognize. It is inversely proportional to the pixel value of the aerial image. The higher the pixel value, the lower the standard segmentation rate. For a single learning iteration, the corresponding standard segmentation rate is the single segmentation rate.

[0060] Standardizing aerial images by segmenting them according to a standard segmentation rate ensures the consistency and standardization of image data. This processing method guarantees that all image data input into the waste disposal model conforms to a certain format and standard, reducing differences caused by factors such as image size and resolution, and improving the stability and accuracy of the waste disposal model. The standard segmentation rate is determined based on the segmentation rate that the waste disposal model can recognize, allowing image data to better adapt to the input requirements of the waste disposal model, thereby improving the model's recognition efficiency and effectiveness.

[0061] By standardizing aerial images and segmenting them into sizes that conform to the requirements of the waste disposal model, unnecessary computation in subsequent processing can be reduced. For example, direct processing of high-resolution aerial images may lead to wasted computing resources and decreased processing speed. Appropriate segmentation allows the image to be decomposed into smaller units for processing, improving efficiency. Standardized image pre-data is also more efficient to store. Segmented image units can be stored more compactly, reducing wasted storage space and facilitating subsequent data management and retrieval.

[0062] The standard segmentation rate is inversely proportional to the pixel count of the aerial image; the higher the pixel count, the lower the standard segmentation rate. This flexible segmentation strategy allows the system to adapt to aerial images of varying qualities, processing both high-resolution and low-resolution images with an appropriate segmentation rate, thus improving the system's flexibility.

[0063] Specifically, in implementation, there are no restrictions on the architecture of the waste disposal model. Existing open-source models that can identify waste disposal regions can be used, or a model that can identify waste disposal regions can be trained independently. For example, a large number of image samples labeled with waste disposal regions can be collected as training samples, and the commonly used U-Net model architecture in the field of image segmentation can be used for training to train a model that can identify waste disposal regions. This will not be elaborated further.

[0064] Through multiple iterations of optimization and testing, the waste disposal model can be trained and validated on different data subsets, thereby reducing errors caused by data bias or overfitting. This rigorous training and testing process improves the reliability of the waste disposal model in practical applications. When the test results meet the preset test results, the learned parameters of the waste disposal model are saved and output as standard parameters, ensuring that the final parameters of the waste disposal model are verified and reliable, and can be used as standard waste disposal model parameters for subsequent monitoring tasks.

[0065] The training process of iterative optimization and testing is highly flexible and can be adjusted according to different monitoring needs and data characteristics. For example, the size of the training dataset can be increased or the testing criteria adjusted to adapt to different application scenarios. The saved standard parameters can serve as the basis for subsequent updates and optimizations of the waste disposal model. With the accumulation of more data and the introduction of new technologies, the waste disposal model can be further optimized to improve the system's scalability and adaptability.

[0066] By iteratively optimizing, testing, and saving the training parameters of the waste disposal model, the accuracy, reliability, and performance optimization of the model can be ensured. This rigorous training process not only improves the performance of the waste disposal model in practical applications but also enhances the system's scalability, stability, and operational efficiency, while reducing operating costs.

[0067] Specifically, the steps for the waste disposal segmentation model to learn from image pre-data include: Adjust the learning parameters of the waste disposal model to standard parameters; The image pre-data is fed into the waste disposal segmentation model; The waste disposal model learns from image pre-data to identify and label waste disposal areas and utilized waste disposal areas.

[0068] In practical implementation, adjusting the learning parameters of the waste disposal model to standard parameters ensures that the model starts from a validated and reliable starting point in each learning process. This avoids performance fluctuations caused by differences in parameter initialization and guarantees the consistency and stability of the model across different learning tasks. By using standard parameters, the waste disposal model can more accurately identify and mark waste disposal areas and utilized waste disposal areas. This step is a crucial link in the entire monitoring system, directly affecting the accuracy of subsequent calculations of waste disposal and utilization areas, as well as the final assessment of the industrial waste disposal interception rate.

[0069] By feeding image pre-data into the waste disposal segmentation model and learning from it, the model can continuously adapt to new data features. Even in different monitoring areas or under different environmental conditions, the model can adjust its parameters through learning, maintaining good generalization ability. During each learning process, the model updates its parameters based on new image pre-data, enabling it to dynamically adapt to changes in the monitoring area and promptly detect new waste disposal areas or changes in utilized waste disposal areas.

[0070] The waste disposal segmentation model can automatically learn from pre-image data and identify and mark waste disposal areas and utilized waste disposal areas. During the learning process, the model can output marking results in real time, providing immediate data support for subsequent area calculations and interception rate assessments. This real-time capability is particularly important for dynamic monitoring, enabling timely detection of potential problems and prompt action.

[0071] After standardization, the pre-data images are better adapted to the waste disposal segmentation model. This allows the model to process data more efficiently during the learning process, reducing errors caused by data quality issues. The automatic labeling function of the waste disposal segmentation model can quickly identify key waste disposal areas and utilized waste disposal areas, reducing unnecessary data redundancy in subsequent processing and further optimizing the data processing workflow.

[0072] By adjusting the learning parameters of the waste disposal model, inputting pre-data images, and labeling them, the consistency and accuracy of the waste disposal model can be ensured, learning efficiency can be improved, the generalization ability of the waste disposal model can be enhanced, the automation level of the system can be increased, the data processing flow can be optimized, and the reliability and stability of the system can be improved.

[0073] The steps for determining the waste disposal area and the utilization area include: Edge detection algorithms are used to extract the boundary contours of the waste disposal area and the utilized waste disposal area; The boundary contour is rendered and fitted to obtain the corresponding closed curve; The area enclosed by the closed curve is calculated using the integral method to obtain the corresponding waste disposal area and utilization area.

[0074] In practical implementation, edge detection algorithms (such as Canny edge detection and the Sobel operator) can accurately extract the boundary contours of the waste disposal area and the utilized waste disposal area. These algorithms can detect areas of pixel intensity variation in the image, thereby accurately delineating the edges of the waste disposal area. Edge detection algorithms can handle waste disposal areas of various complex shapes, accurately extracting boundary contours regardless of their irregularity. This is crucial for subsequent area calculations, as complex boundary shapes can lead to errors in area calculations.

[0075] Rendering and fitting of boundary contours: Smoothing Boundary Curves: The extracted boundary contours are rendered and fitted, which fits discrete edge points into smooth, closed curves. This step eliminates noise and small irregularities in the boundary contours, making the boundaries smoother and more continuous, thereby improving the accuracy of area calculations.

[0076] By fitting closed curves, waste disposal areas of different scales can be better handled. Regardless of the area size, the fitted curve can accurately describe the boundary of the area, providing a reliable basis for subsequent area calculations.

[0077] High-precision area calculation: Calculating the area enclosed by a closed curve using integration methods is a mathematically precise approach. This method can handle closed regions of arbitrary shapes, regardless of their complexity, and accurately calculates their area. For example, Green's Theorem can be used to calculate the area enclosed by polygons or curves. By accurately extracting boundary contours, smoothing the fit, and performing high-precision area calculations, more accurate data on waste disposal and utilization areas can be provided. This data is the foundation for calculating industrial waste interception rates, and accurate area data improves the accuracy of interception rate calculations. By fitting closed curves, complex boundary contours can be simplified into a set of parameterized curves, reducing the need for data storage. This not only saves storage space but also facilitates subsequent data management and retrieval.

[0078] By using an edge detection algorithm to extract boundary contours, rendering and fitting the boundary contours, and using an integral method to calculate the area, the boundaries between the waste disposal area and the utilized waste disposal area can be accurately extracted, improving the accuracy of area calculation, enhancing the automation and flexibility of the system, and reducing system operating costs.

[0079] Specifically, the industrial waste interception rate of the monitoring area is calculated based on the area of ​​waste disposal and the area of ​​utilization. The industrial waste interception rate is the quotient of the waste area and the utilization area.

[0080] Specifically, the industrial waste interception rate is compared with the interception rate threshold. When the industrial waste interception rate is greater than the interception rate threshold, the monitoring area is not adjusted.

[0081] Specifically, when the industrial waste interception rate is less than the interception rate threshold, a sedimentation tank is set up in the monitoring area, wherein... Sedimentation basins are used to guide industrial waste from the waste disposal area to the waste disposal area that has been utilized.

[0082] In practical implementation, the interception rate is quantified: the industrial waste interception rate is obtained by calculating the quotient of the waste disposal area and the utilization area. This quantitative method can intuitively reflect the effectiveness of waste interception and provide data support for subsequent decision-making. Regularly calculating the industrial waste interception rate allows for dynamic monitoring of changes in waste interception effectiveness, timely identification of potential problems, and real-time feedback for soil and water conservation.

[0083] The industrial waste interception rate is compared with a preset interception rate threshold, and the result determines whether a sedimentation tank needs to be set up in the monitoring area. This data-driven decision-making method is more scientific and objective, avoiding errors from subjective judgment.

[0084] When the industrial waste interception rate exceeds the threshold, it indicates that the current interception effect meets the requirements, and no additional investment is needed. This helps optimize resource utilization and avoid unnecessary engineering construction.

[0085] When the industrial waste interception rate is lower than the threshold, it indicates that the waste interception effect is unsatisfactory and measures need to be taken. By setting up sedimentation ponds in the monitoring area, waste can be effectively guided from the waste disposal area to the utilized waste disposal area, reducing soil erosion. Sedimentation ponds can also reduce pollution of downstream water bodies by waste, protecting the ecological environment and achieving sustainable development. The entire calculation and comparison process can be automated, reducing the need for manual intervention. This not only improves work efficiency but also reduces problems caused by human error.

[0086] When the industrial waste interception rate is less than the interception rate threshold, the system can automatically trigger the decision to set up a sedimentation tank, further improving the system's automation level.

[0087] By calculating the industrial waste interception rate and making decisions based on the comparison results, the effectiveness of waste interception can be accurately assessed, enabling scientific decision-making, effectively controlling soil erosion, and making the entire monitoring system more efficient, accurate, and economical, thus better meeting the needs of dynamic monitoring of soil erosion.

[0088] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

[0089] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for dynamic monitoring of soil erosion using remote sensing based on unmanned aerial vehicles (UAVs), characterized in that, include: The drone is used to collect aerial images of the monitoring area according to a preset flight path; Basic landform visual features are extracted from aerial images acquired within a predetermined time domain; Identify salient boundaries in aerial images, construct reference windows based on salient boundaries, determine local reference windows segmented by salient boundaries, and enhance local images within local reference windows; By comparing the enhanced local image with the basic terrain visual features, the aberrant local reference window is locked, and the aberrant local reference window is verified based on the sample features to determine whether the aerial image should be marked. The analysis of the marked aerial images includes using a waste disposal segmentation model to identify and segment the waste disposal area and the utilized waste disposal area in the aerial images in order to determine the corresponding waste disposal area and utilization area. Calculate the industrial waste interception rate in the monitoring area, and determine whether to set up a sedimentation tank in the monitoring area based on the comparison result between the industrial waste interception rate and the interception rate threshold.

2. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 1, characterized in that, The process of extracting basic landform visual features from aerial images acquired within a predetermined time domain includes, Extract the chromaticity values ​​and texture density from each frame of aerial images within a predetermined time domain; The mean chromaticity value and the mean texture density are calculated as the basic visual features of the terrain.

3. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 2, characterized in that, The process of identifying salient boundaries in aerial images, constructing reference windows based on these salient boundaries, and determining the local reference windows segmented by the salient boundaries includes: The contour lines in the aerial image are determined. If the difference in chromaticity values ​​on both sides of the contour line is greater than a predetermined chromaticity difference threshold, the contour line is determined as a significant boundary. A reference window is overlaid on a significant boundary such that the significant boundary passes through the reference window, thereby dividing the reference window into local reference windows.

4. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 3, characterized in that, The process of locking down anomalous local reference windows by comparing enhanced local images with basic terrain visual features includes the following steps. Extract the visual features of the terrain within the local reference viewports; Compare the visual features of landforms with the visual features of basic landforms; If the condition of topographic difference is met, the local reference window is determined to be an aberrant local reference window; Among them, the landform difference condition is that the chromaticity difference ratio is greater than a predetermined chromaticity difference ratio threshold and the texture density difference ratio is greater than a predetermined texture density difference ratio threshold.

5. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 1, characterized in that, The process of determining whether to tag aerial images includes, Obtain sample features, including sample chromaticity values ​​and sample texture density; The mutation reference window is validated, including extracting the chromaticity value and texture density value within the mutation reference window; Determine the chromaticity difference ratio of the chromaticity value relative to the sample chromaticity value, and determine the texture density difference ratio of the texture density value relative to the sample texture density value; If the chromaticity difference ratio is less than a preset sample feature difference ratio threshold and the texture density difference ratio is less than a preset sample texture feature difference ratio threshold, then it is determined that the aerial image containing the variant reference window should be marked.

6. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 5, characterized in that, The steps of the waste disposal segmentation model in learning from image pre-data include: The learning parameters of the waste disposal model were adjusted to standard parameters; The image pre-data is then passed into the waste disposal segmentation model; The waste disposal segmentation model learns from the image pre-data, identifies the waste disposal area and the utilized waste disposal area, and marks them.

7. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 6, characterized in that, The steps for determining the waste disposal area and the utilization area include: The boundary contours of the waste disposal area and the utilized waste disposal area are extracted using an edge detection algorithm. The boundary contour is rendered and fitted to obtain the corresponding closed curve; The area enclosed by the closed curve is calculated using an integral method to obtain the corresponding waste area and utilization area.

8. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 7, characterized in that, The industrial waste interception rate of the monitoring area is calculated based on the waste disposal area and the utilization area, wherein... The industrial waste interception rate is the quotient of the waste area and the utilization area.

9. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 8, characterized in that, The industrial waste interception rate is compared with the interception rate threshold. When the industrial waste interception rate is greater than the interception rate threshold, the monitoring area is not adjusted.

10. The method for dynamic monitoring of soil erosion based on UAV remote sensing according to claim 9, characterized in that, When the industrial waste interception rate is less than the interception rate threshold, a sedimentation tank is installed in the monitoring area. The sedimentation tank is used to guide industrial waste from the waste disposal area to the utilized waste disposal area.

Images