Submerged vegetation coverage monitoring and suitable recovery area rapid identification method

Abstract

The invention provides a submerged vegetation coverage monitoring and suitable recovery area rapid identification method, and relates to the field of water ecology protection and restoration, and the method comprises the steps: collecting underwater images, and forming an underwater image data set; an image classification model is adopted to classify the underwater images, and submerged vegetation features, coverage information and sediment type information of each monitoring point site are identified and obtained; carrying out spatial interpolation processing by adopting a spatial interpolation algorithm and utilizing the submerged vegetation characteristics and coverage information, bottom mud type information and water depth data of each monitoring point position, and generating a water depth, bottom mud type, submerged vegetation type and coverage spatial distribution map of the target water area; and utilizing the spatial distribution map, performing evaluation through an ecological suitability comprehensive evaluation method, and determining a submerged vegetation recovery type, a suitable recovery area and a corresponding recovery priority of the target water area. According to the method, the characteristics of the submerged plants and the bottom mud in the water body can be rapidly monitored, the recovery suitable area of the submerged plants is accurately identified, and the recovery priority time sequence is reasonably arranged.

Description

Technical Field

[0001] The present invention relates to the field of water ecological protection and restoration, and in particular to a method for monitoring submerged vegetation coverage and quickly identifying suitable restoration areas. Background Art

[0002] Submerged vegetation, as aquatic plant communities that grow at the bottom of water bodies, with their roots fixed to the bottom mud and their stems and leaves completely or partially immersed in water, is an indispensable component of aquatic ecosystems. By constructing complex underwater structures, they provide important habitats, reproduction and shelter for various aquatic animals, such as serving as spawning substrates for fish and attachment interfaces for plankton. At the same time, submerged vegetation plays a key ecological function in improving water quality (such as absorbing nutrients and increasing water transparency), inhibiting the excessive reproduction of harmful algae (such as algal blooms), stabilizing bottom mud and reducing resuspension, and is essential for maintaining aquatic biodiversity and the health of ecosystems.

[0003] However, with the intensification of human activities, many lakes and other water bodies are facing severe eutrophication, leading to a rapid degradation and even extinction of submerged vegetation. To address this challenge, large-scale artificial restoration projects for submerged vegetation have been widely carried out both domestically and internationally. Their core goal is to rebuild submerged plant communities, enhance the self-purification capacity of water bodies, effectively control algal blooms, thereby improving overall water quality and enhancing the species diversity and structural stability of aquatic ecosystems. Traditional preliminary work on vegetation restoration relies primarily on manual methods to assess the status of submerged vegetation and sediment conditions. For example, vegetation monitoring often relies on visual estimation or sampling and analysis using tools such as grass clips to determine cover and species composition. Sediment surveys typically use mud buckets or columnar samplers to obtain samples, which are then analyzed in the laboratory for physical and chemical properties such as moisture content and particle size distribution, as well as pollutant loads, to determine whether the target area meets the basic conditions for vegetation restoration. These traditional methods have significant limitations: not only are they difficult to achieve comprehensive coverage of the entire water area, they are also labor-intensive and resource-intensive, inefficient, and can disturb or even damage fragile submerged plants and their bottom habitats during sampling. More importantly, traditional methods often fail to fully account for the complex spatial heterogeneity of water depth, the distribution of existing vegetation patches, and sediment conditions, as well as their inter-ecological relationships. This often results in a lack of scientific basis and precise spatial positioning for the selection of restoration areas.

[0004] The invention patent application in China, CN202110973730.X, discloses a rapid monitoring system and control method for submerged plant communities. This technical solution uses shipboard equipment to collect data such as water transparency, nutrients, water depth, and submerged plant images in real time. After analysis by the data processing module, the control device links the water conservancy facilities to automatically adjust the water depth or nutrient concentration, aiming to achieve dynamic monitoring and control of the existing submerged plant growth environment. This system effectively improves the level of automated management of the growth status of existing submerged plant communities. However, it mainly focuses on maintaining and optimizing the growth conditions of existing vegetation areas, and does not provide targeted technical means for how to systematically identify and evaluate areas with restoration potential in wider waters that may lack vegetation. In the early stages of vegetation restoration, how to effectively conduct large-scale suitability assessments and site selection remains a technical problem that needs further resolution. Summary of the Invention

[0005] In view of this, the present invention provides a method for monitoring submerged vegetation coverage and quickly identifying suitable restoration areas, which can quickly monitor the characteristics of submerged plants and bottom mud in water bodies, accurately identify suitable areas for submerged plant restoration and reasonably arrange restoration priority sequences.

[0006] The technical solution of the present invention is achieved as follows:

[0007] The present invention provides a method for monitoring submerged vegetation coverage and quickly identifying suitable restoration areas, comprising:

[0008] S1. Use underwater image acquisition equipment to collect underwater image sequences in the target waters along a preset route, and simultaneously record the geographic coordinates and water depth data corresponding to the underwater images to form an underwater image dataset with geographic coordinates and water depth information;

[0009] S2. Use an image classification model to process and classify the underwater image dataset to identify the submerged vegetation type, coverage information, and sediment type information at each monitoring point. Manually verify the low-confidence classification results using a composite confidence level, and use the verification results as feedback to optimize the model parameters of the image classification model.

[0010] S3. Using a spatial interpolation algorithm, using the submerged vegetation type, coverage information, sediment type information, and water depth data of each monitoring point, perform spatial interpolation processing to generate a spatial distribution map of the target water area;

[0011] S4. Utilize the spatial distribution map of the target waters, combined with preset ecological evaluation factors, and conduct a quantitative assessment through the comprehensive ecological suitability evaluation method to determine the types of submerged vegetation restoration in the target waters, suitable restoration areas, and corresponding restoration priorities.

[0012] Preferably, in step S1, the underwater image acquisition equipment includes a high-definition underwater camera, a depth sounder and a GPS locator installed on a cruise boat; the preset route is a spiral or Z-shaped route; the underwater image acquisition equipment moves along the route and collects data.

[0013] Preferably, in step S2, the image classification model is an adaptive multi-threshold underwater image enhancement and classification integrated model, which sequentially performs a series of image processing operations on the underwater image, including image enhancement preprocessing, multi-scale boundary enhancement processing, and adaptive classification processing, to obtain preliminary classification labels for submerged vegetation at each monitoring point and preliminary classification labels for bottom sediment.

[0014] Preferably, the composite confidence is used to evaluate the preliminary classification labels, and the calculation method is:

[0015] Conf(p)=γ1·P max (p)+γ2·(1-Entropy(p))+γ3·S spatial (p)

[0016] Among them, Conf(p) is the classification confidence score of point p; P max (p) is the highest category probability value in the classification result corresponding to point p; Entropy(p) is the normalized probability distribution entropy of the classification result corresponding to point p; S spatial (p) is the spatial consistency score between the classification result of point p and the classification result of its neighborhood; γ1, γ2, γ3 are preset weight coefficients, satisfying γ1+γ2+γ3=1;

[0017] If the Conf(p) of point p is lower than the preset confidence threshold Conf threshold , then the preliminary classification label of the point is marked as a label to be manually verified.

[0018] Preferably, in step S3, the spatial interpolation algorithm performs spatial interpolation processing on the submerged vegetation type, coverage information, sediment type information and water depth data of each monitoring point by integrating the spatial correlation structure model of hydrodynamic correction and the ecological threshold constraint conditions, and combines the feature boundary maintaining interpolation adjustment to generate a spatial distribution map of the target water area, including a spatial distribution map of submerged vegetation coverage, a spatial distribution map of sediment type and a spatial distribution map of water depth.

[0019] Preferably, the hydrodynamically corrected spatial correlation structure model is a model using a preset hydrodynamic correction function After correcting the semivariogram function γ0(h), the relationship is:

[0020]

[0021] Where γ(h) is the modified semivariogram function; h is the distance vector between two points in space; is the dominant water flow direction vector; θ is the water flow intensity parameter.

[0022] Preferably, the spatial distribution map generated in step S3 also includes: a spatial distribution map of vegetation species composition and a distribution map of sediment health status.

[0023] Preferably, in step S4, the ecological suitability comprehensive evaluation method is to use the integrated optimized ecological suitability comprehensive evaluation index IECSI for calculation, and the calculation formula of IECSI is:

[0024]

[0025] Among them, IECSI(p) is the comprehensive ecological suitability index of point p in the target water area; is the standardized score of the i-th core restriction factor at point p; is the preset weight index of the i-th core limiting factor, satisfying K is the number of core limiting factors; C neighbor (p) is the connectivity index of the neighboring vegetation at point p; α is the neighboring gain adjustment coefficient; is the standardized score of the jth auxiliary support factor at point p; is the preset weight coefficient of the jth auxiliary support factor, satisfying M is the number of auxiliary support factors; Ψ(p) is the timing stability correction coefficient at point p.

[0026] Preferably, the core limiting factors include historical vegetation distribution, water depth suitability and sediment health; the auxiliary supporting factors include sediment particle size suitability and sediment moisture content suitability.

[0027] Preferably, the method further comprises:

[0028] Based on the comprehensive evaluation index value of ecological suitability at each point, the target water area is divided into priority restoration areas, suboptimal restoration areas, areas to be observed, and unsuitable areas according to the preset index threshold range. The submerged vegetation restoration types and suitable restoration area classification maps of the target water area are output;

[0029] In areas where submerged vegetation restoration measures have been implemented, steps S1, S2, and S3 are repeated periodically to obtain the submerged vegetation type, coverage information, sediment type information, and corresponding spatial distribution map of the restored water area. The monitoring results are compared and analyzed with the preset restoration targets, and subsequent restoration strategies or monitoring plans are adjusted based on the analysis results.

[0030] The present invention has the following beneficial effects compared to the prior art:

[0031] (1) The present invention can efficiently and accurately obtain multi-source spatial data such as submerged vegetation, sediment characteristics, and water depth in the target waters, and conduct comprehensive evaluation and zoning based on ecological principles. This provides strong technical support for the scientific site selection and priority determination of restoration areas in water ecological restoration projects, thereby improving the scientific nature and decision-making efficiency of submerged vegetation restoration work.

[0032] (2) The present invention uses underwater image acquisition equipment equipped with a high-definition underwater camera, a depth sounder, and a GPS locator, and performs automated and synchronized data acquisition along a preset route. Compared with traditional manual sampling and fixed-point monitoring methods, it greatly improves the efficiency of data acquisition and the integrity of spatial coverage. At the same time, the vegetation images, water depth, and geographic coordinate data obtained have a precise temporal and spatial correspondence, and effectively avoids human disturbance of the underwater sediment and vegetation habitat.

[0033] (3) The present invention adopts an adaptive multi-threshold underwater image enhancement and classification integrated model, combined with a composite confidence assessment and necessary manual verification feedback optimization mechanism, which can automatically and finely identify the submerged vegetation type, coverage and sediment type of the collected underwater image data, effectively reducing the subjectivity and tediousness of traditional manual interpretation and improving the objectivity and reliability of the recognition results;

[0034] (4) By using a spatial interpolation algorithm, the present invention converts discrete monitoring point data into a continuous spatial distribution map. Furthermore, combined with a comprehensive ecological suitability evaluation method, it can systematically reveal the spatial heterogeneity of key environmental factors for submerged vegetation growth and their coupling relationship with vegetation distribution, thereby scientifically and quantitatively identifying areas with different restoration potentials, providing an intuitive spatial decision-making basis for the formulation of submerged vegetation restoration plans.

[0035] (5) The present invention can not only quickly identify suitable restoration areas for submerged vegetation, but also divide the restoration priorities of target waters by establishing a classification system for submerged vegetation restoration types and suitable restoration areas, and support periodic dynamic evaluation and feedback optimization of the effects of areas where restoration measures have been implemented, thus forming a closed-loop technical process from early site selection, mid-term planning to later management and adaptive adjustment, which helps to improve the long-term effectiveness and sustainability of submerged vegetation restoration projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0037] Figure 1 is a flow chart of the method of the present invention;

[0038] Figure 2 It is a technical implementation diagram of the present invention;

[0039] Figure 3 This is a schematic diagram of unsupervised classification for identifying submerged vegetation in water according to an embodiment of the present invention;

[0040] Figure 4 This is a schematic diagram of supervised classification and identification of sediment samples according to another embodiment of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0042] like Figure 1 As shown, the present invention provides a method for monitoring submerged vegetation coverage and quickly identifying suitable restoration areas, comprising:

[0043] S1. Use underwater image acquisition equipment to collect underwater image sequences in the target waters along a preset route, and simultaneously record the geographic coordinates and water depth data corresponding to the underwater images to form an underwater image dataset with geographic coordinates and water depth information;

[0044] S2. Use an image classification model to process and classify the underwater image dataset to identify the submerged vegetation type, coverage information, and sediment type information at each monitoring point. Manually verify the low-confidence classification results using a composite confidence level, and use the verification results as feedback to optimize the model parameters of the image classification model.

[0045] S3. Using a spatial interpolation algorithm, using the submerged vegetation type, coverage information, sediment type information, and water depth data of each monitoring point, perform spatial interpolation processing to generate a spatial distribution map of the target water area;

[0046] S4. Utilize the spatial distribution map of the target waters, combined with preset ecological evaluation factors, and conduct a quantitative assessment through the comprehensive ecological suitability evaluation method to determine the types of submerged vegetation restoration in the target waters, suitable restoration areas, and corresponding restoration priorities.

[0047] like Figure 2As shown, the present invention proposes a method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas. By systematically integrating underwater data acquisition, intelligent image processing, and spatial ecological analysis technologies, a complete technical chain of "data acquisition - intelligent classification - manual verification - spatial modeling - ecological evaluation" is constructed. Specifically, the method first uses a cruise vessel, such as an unmanned vessel, equipped with a high-definition underwater camera, a depth sounder, and a GPS locator to collect underwater image sequences and water depth data with geographic coordinates along a preset route. An adaptive multi-threshold underwater image enhancement and classification integrated model is then used to process and classify the collected underwater images, identifying the submerged vegetation cover and sediment type at each monitoring point. Low-confidence points are screened for manual verification using a composite confidence level, forming a feedback optimization closed loop. A spatial interpolation algorithm is then used to integrate a hydrodynamically corrected spatial correlation structure model and ecological threshold constraints to transform the discrete monitoring data into a continuous spatial distribution map. Finally, based on the generated spatial distribution map, a fused and optimized comprehensive ecological suitability evaluation index (IECSI) is used. This IECSI comprehensively considers core limiting factors (such as historical vegetation distribution, water depth suitability, and sediment health) and auxiliary supporting factors (such as sediment particle size, moisture content, water quality, and light conditions) to scientifically and quantitatively assess the ecological suitability of each point. This allows for the determination of submerged vegetation restoration types, suitable restoration areas, and their restoration priorities, supporting dynamic evaluation and feedback optimization of restoration outcomes. The present invention effectively improves the efficiency of submerged vegetation monitoring and the scientific nature of restoration area site selection through this data-driven, multi-factor fusion technical method.

[0048] Specifically, in one embodiment of the present invention, step S1 includes:

[0049] The underwater image acquisition equipment consists of a cruising boat, such as an unmanned vessel, equipped with a high-definition underwater camera, a depth sounder, and a GPS locator. The HD underwater camera, mounted vertically downward from the bottom of the boat, captures underwater images; the depth sounder obtains water depth information corresponding to the image location; and the GPS locator records the precise geographic coordinates of the sampling point. These three devices work in sync to ensure that each underwater image is accurately mapped to its corresponding geographic location and depth.

[0050] The default route is a spiral or zigzag route:

[0051] Spiral route: Suitable for larger open water areas, such as the central area of ​​a lake. This route starts from the center of the target water area and expands outward in a spiral shape. The route spacing can be adjusted based on the water transparency and target monitoring accuracy.

[0052] Zigzag Route: Suitable for narrow waterways or bays. This route consists of parallel segments connected by short distances to form a continuous "Z"-shaped path.

[0053] The underwater image acquisition device moves along a preset route and collects data to form an underwater image dataset. In this embodiment, after the underwater image is collected, preliminary data preprocessing can be performed, including optical correction, scale calibration, and coordinate association. A unified point attribute table is established, including:

[0054] Serial number (ID): unique identification code; spatial coordinates (X, Y): plane coordinates or longitude and latitude; water depth (D): water depth at the sampling point, unit: m; image file association (Image_ID): corresponding underwater image file number; other environmental parameters: water quality parameters, sediment type, etc.

[0055] Specifically, in one embodiment of the present invention, in step S2, the image classification model is an adaptive multi-threshold underwater image enhancement and classification integrated model, which sequentially performs a series of image processing operations on the underwater image, including image enhancement preprocessing, multi-scale boundary enhancement processing, and adaptive classification processing, to obtain preliminary classification labels for submerged vegetation at each monitoring point and preliminary classification labels for bottom sediment.

[0056] S2.1 Underwater Image Enhancement Preprocessing

[0057] In order to solve the imaging problem under different water depths and water quality conditions, an adaptive multi-threshold underwater image enhancement algorithm is designed:

[0058] I enhanced (x,y)=I original (x,y)·e α(d)·(1-β·T(x,y))

[0059] Among them, I enhanced (x, y) is the enhanced image pixel value; I original (x, y) is the original image pixel value; α(d) is the compensation coefficient related to the water depth d, α(d) = k1·ln(d+1), k1 is the adjustment parameter (1.0-2.0); β is the light attenuation adjustment parameter (0.3-0.7), which is automatically adjusted according to the turbidity of the water body; T(x, y) is the transmittance estimate, calculated using the dark channel prior method.

[0060] S2.2 Multi-scale boundary enhancement processing

[0061] To address the issue of blurred boundaries between vegetation and sediment, multi-scale boundary enhancement processing is applied:

[0062]

[0063] in, It is the result of multi-scale edge enhancement; For different scales (σ i )’s Gaussian kernel; w i is the weight coefficient of each scale, satisfying K is the number of scales, usually 3-5; * is the convolution operator.

[0064] S2.3 Adaptive classification strategy selection

[0065] Automatically select the best classification strategy based on image characteristics:

[0066]

[0067] Among them, E(I) is the image entropy value, which measures the information complexity; V(I) is the image variance, which measures the contrast level; M(I) is the prior model matching score; τ1, τ2, τ3 are adaptive thresholds, which are dynamically adjusted according to historical data.

[0068] S2.4 Unsupervised classification method (submerged vegetation identification)

[0069] When unsupervised classification is selected, the improved K-means or ISODATA algorithm with multi-feature fusion is used:

[0070] Extract color feature vectors: hue, saturation, and lightness in HSV space. Extract texture feature vectors: contrast, correlation, energy, and homogeneity from the gray-level co-occurrence matrix. Normalize and fuse feature vectors. Execute an improved clustering algorithm to automatically determine the optimal number of clusters. Map clustering results to vegetation categories using expert rules or sample library matching.

[0071] S2.5 Supervised Classification Method (Sediment Health Status Assessment)

[0072] When choosing supervised classification, use an ensemble learning framework:

[0073]

[0074] Among them, P(y i |x) is the sample x belonging to category y i The comprehensive probability of j (y i |x) is the probability given by the j-th classifier; ω j is the weight of the j-th classifier, satisfying M is the number of classifiers (3-5).

[0075] The classifier weights are automatically adjusted through periodic evaluation:

[0076]

[0077] Acc j is the accuracy of the j-th classifier on the manual verification sample set.

[0078] It should be noted that when performing image classification, it is not necessary to select only one classification method. The classification strategy described in this embodiment can refer to selecting a classification method at different stages. For example, color extraction can be used to identify whether submerged plants are present. If so, unsupervised classification can be used to identify the submerged plant type. After subsequent manual verification, supervised classification can be further used to improve the classification results and update the attribute table. Alternatively, a supervised classification algorithm can be used, which requires pre-training with training samples before identification to obtain the classification results.

[0079] S2.6 Composite confidence

[0080] In order to reduce the workload of manual review, a composite confidence level is designed:

[0081] Conf(p)=γ1·P max (p)+γ2·(1-Entropy(p))+γ3·S spatial (p)

[0082] Among them, Conf(p) is the classification confidence score of point p; P max (p) is the highest category probability value in the classification result corresponding to point p; Entropy(p) is the normalized probability distribution entropy of the classification result corresponding to point p; S spatial (p) is the spatial consistency score between the classification result of point p and the classification result of its neighborhood; γ1, γ2, γ3 are preset weight coefficients, satisfying γ1+γ2+γ3=1;

[0083] in, N(p) is the set of neighboring points of point p; w(p,q) is the weight between point p and neighboring point q, calculated based on distance; δ(y(p), y(q)) is the Kronecker delta function, which is 1 when the classification results of p and q are the same, and 0 otherwise.

[0084] If the Conf(p) of point p is lower than the preset confidence threshold Conf threshold , then the preliminary classification label of the point is marked as a label to be manually verified.

[0085] For points whose automatic classification confidence is lower than the threshold, manual verification is performed: the original underwater image is extracted and the automatic classification results are displayed; experts interpret and confirm the actual situation of the point; the point attribute table is updated and the classification results are corrected; the manual verification results are fed back to the classification algorithm to adjust the model parameters.

[0086] After the above processing, the submerged vegetation type, coverage information and sediment type information of each monitoring point are finally output, including: submerged vegetation coverage, submerged vegetation type, sediment type, sediment health status, and confidence score of each classification result.

[0087] Specifically, in one embodiment of the present invention, in step S3, the spatial interpolation algorithm integrates the hydrodynamically corrected spatial correlation structure model and the ecological threshold constraint conditions to perform spatial interpolation processing on the submerged vegetation type, coverage information, sediment type information and water depth data of each monitoring point, and combines the feature boundary-preserving interpolation adjustment to generate a spatial distribution map of the target water area, including a submerged vegetation coverage spatial distribution map, a sediment type spatial distribution map and a water depth spatial distribution map.

[0088] In this embodiment, based on the updated attribute table, the spatial interpolation algorithm ECAI is used to generate various distribution maps:

[0089] S3.1 Hydrodynamically assisted spatially correlated structure modeling

[0090] Traditional spatial interpolation methods usually assume that spatial correlation is only affected by distance, ignoring the impact of water flow, an important hydrodynamic factor, on the spatial distribution of aquatic ecosystems. In the ECAI algorithm, the hydrodynamic factor is introduced to modify the semivariogram function:

[0091]

[0092] Among them, γ(h) is the modified semivariogram function; γ0(h) is the original semivariogram function (spherical model, exponential model, etc.); is the hydrodynamic correction function; is the dominant water flow direction vector; θ is the water flow intensity parameter.

[0093] The hydrodynamic correction function is defined as:

[0094]

[0095] in, is the distance vector h and the direction of water flow ; δ is the directional strength parameter (0.2-0.8); |h| is the modulus of the distance vector.

[0096] The preset hydrodynamic correction function has the following characteristics: when the distance vector is consistent with the water flow direction (the angle is 0°), The correction effect is the strongest; when the distance vector is perpendicular to the water flow direction (the angle is 90°), No directional correction; when the distance vector is opposite to the direction of the water flow (the angle is 180°), Reduce correlation; as the distance increases, the correction effect decays exponentially, reflecting the distance attenuation characteristics of water flow influence.

[0097] S3.2 Ecological Threshold Constrained Kriging Interpolation

[0098] Traditional Kriging interpolation may produce ecologically unreasonable interpolation results. To address this problem, the ECAI algorithm introduces ecological threshold constraints and incorporates ecological knowledge into the Kriging interpolation process:

[0099]

[0100] where the weight λ i Satisfy the following ecological constraints:

[0101]

[0102] Among them, [a k ,b k ] is the region R k The ecological threshold interval of R k sub-regions with similar ecological conditions.

[0103] These constraints ensure that the interpolation results do not produce ecologically unreasonable extreme values, such as impossible vegetation cover or sediment distribution that does not conform to physical laws. For example, for submerged vegetation cover, its value range is guaranteed to be between 0 and 100%. For specific areas, the upper limit of vegetation cover is set based on environmental conditions (such as water depth and light). For sediment type, the interpolation results are ensured to conform to the natural transition pattern of sediment distribution.

[0104] S3.3 Multivariate environmental factor information entropy discrimination partitioning

[0105] According to environmental factors such as water depth and bottom type, spatial information entropy is discriminated and partitioned:

[0106]

[0107] Where E(R) is the normalized information entropy of region R; p i is the normalized probability density of environmental factor i in the region; m is the number of environmental factors considered.

[0108] When E(R)>E threshold When the area is further subdivided, the ecological homogeneity of the sub-regions is ensured.

[0109] Specifically, environmental factors may include water depth gradient, bottom type, distance from shoreline, and water quality parameters.

[0110] S3.4 Feature boundary preserving interpolation

[0111] Ecosystems often have distinct boundary features, such as the junction of different substrate types and areas with rapid changes in water depth. To maintain these ecological boundary features, the ECAI algorithm uses boundary-preserving interpolation adjustments:

[0112] ZBP (s)=Z K (s)·(1-B(s))+Z E (s)·B(s)

[0113] Among them, Z BP (s) is the boundary-preserving interpolation result; Z K (s) is the Kriging interpolation result; Z E (s) is the local interpolation result of the boundary area; B(s) is the boundary detection function (0-1).

[0114] The boundary detection function is defined as:

[0115]

[0116] Where d(s,E) is the distance from point s to the nearest boundary E; σ B is the boundary influence attenuation parameter; r B The border influence radius.

[0117] In summary, the complete implementation process of the ECAI algorithm of this embodiment is as follows: 1) sorting the point attribute table obtained in step S2, including the submerged vegetation type, coverage information, sediment type information and water depth data of each monitoring point, obtaining the water area hydrodynamic parameters, such as the dominant water flow direction vector and water flow intensity parameter, and initializing various interpolation parameters, such as the directional intensity parameter and the boundary influence parameter; 2) calculating the spatial distribution characteristics of each environmental factor, applying the multivariate environmental factor information entropy discrimination method to perform spatial partitioning, and determining the ecological threshold constraint interval for each sub-area [a k ,b k ]; 3) Identify the boundary position of each environmental factor, construct the boundary detection function B(s), and design a local interpolation strategy for the boundary area; 4) Fit the original semivariogram function γ0(h) based on the sample point data and apply the hydrodynamic correction function Calculate the modified semivariogram function γ(h). 5) For each prediction point s0, determine the subregion R to which it belongs k , apply the ecological threshold constraint condition to solve the optimal weight λ i , calculate the Kriging interpolation result 6) For points close to the boundary, calculate the boundary detection function value B(s) and calculate the local interpolation result Z of the boundary area E (s), fused to obtain the final boundary-preserving interpolation result Z BP (s); 7) Generate spatial distribution maps of target waters, including: spatial distribution maps of submerged vegetation coverage, spatial distribution maps of sediment types, spatial distribution maps of water depth, spatial distribution maps of vegetation species composition, and spatial distribution maps of sediment health status.

[0118] Specifically, in one embodiment of the present invention, step S4 determines the submerged vegetation restoration types, suitable restoration areas and restoration priorities of the target waters through a comprehensive evaluation method based on the spatial distribution map obtained in the previous steps and the preset ecological evaluation factors.

[0119] This example constructs an integrated and optimized ecological suitability comprehensive evaluation index (IECSI) based on the generated spatial distribution map to identify suitable restoration areas:

[0120]

[0121] Among them, IECSI(p) is the comprehensive ecological suitability index of point p in the target water area; is the standardized score of the i-th core restriction factor at point p; is the preset weight index of the i-th core limiting factor, satisfying K is the number of core limiting factors; C neighbor (p) is the connectivity index of the neighboring vegetation at point p; α is the neighboring gain adjustment coefficient; is the standardized score of the jth auxiliary support factor at point p; is the preset weight coefficient of the jth auxiliary support factor, satisfying M is the number of auxiliary support factors; Ψ(p) is the timing stability correction coefficient at point p.

[0122] Core limiting factors include historical vegetation distribution, water depth suitability, and sediment health. The specific evaluation methods are as follows:

[0123] Historical vegetation distribution evaluation function:

[0124]

[0125] Water depth suitability evaluation function:

[0126]

[0127] Among them, μ D is the center value of the optimal water depth; σ D is the standard deviation parameter of the suitable water depth range; (D(p)) is the interval indicator function, when D(p)∈[D min ,D max ] is 1 when the value is 0, otherwise it is 0.

[0128] Sediment health status evaluation function:

[0129]

[0130] Among them, C j(p) is the concentration of the jth pollutant at point p; C j,ref is the reference concentration threshold of the jth pollutant; M is the number of pollutant types considered.

[0131] Calculation method of vegetation connectivity index:

[0132]

[0133] Where N(p) is the search neighborhood of point p (e.g., a 500-meter radius circle); V(q) is the vegetation coverage of point q (0-1); d(p,q) is the distance between points p and q; σ d is the spatial influence attenuation parameter.

[0134] Auxiliary supporting factors include suitability of sediment particle size and suitability of sediment moisture content.

[0135] Evaluation function of sediment particle size suitability:

[0136]

[0137] Among them, G(p) is the sediment particle size at point p; G opt,min ,G opt,max is the optimal particle size range for target vegetation; G max is the maximum particle size allowed.

[0138] Evaluation function of suitability of sediment moisture content:

[0139]

[0140] Where W(p) is the moisture content of the bottom mud at point p; W opt is the optimum moisture content of the target vegetation; W max is the maximum allowable moisture content.

[0141] Considering seasonal and interannual fluctuations in hydrological conditions:

[0142]

[0143] Among them, CV T (p) is the time variation coefficient of the water level at point p; ΔW max (p) is the maximum water level fluctuation; W avg (p) is the average water level; β, γ are adjustment parameters (0-0.5).

[0144] According to the calculated IECSI index value, step S4 finally outputs the submerged vegetation restoration type, suitable restoration area and restoration priority of the target water area.

[0145] Specifically, according to the preset index threshold range, the target waters are divided into the following categories:

[0146] Priority restoration area: IECSI(p)≥IECSI high ; Suboptimal recovery area: IECSI medium ≤IECSI(p) <IECSI high ; Observation area: IECSI low ≤IECSI(p) <IECSI medium ; Unsuitable area: IECSI (p) <IECSI low .

[0147] Among them, IECSI low IECSI medium and IECSI high is a predefined threshold.

[0148] Considering spatial continuity, adjacent high-priority areas are merged to form continuous restoration patches. A suitable restoration area map and a restoration priority zoning map are generated.

[0149] Specifically, in one embodiment of the present invention, in areas where submerged vegetation restoration measures have been implemented, a monitoring feedback mechanism needs to be established: fixed monitoring points are set up in the restoration area, and steps S1, S2, and S3 are periodically repeated to obtain: the submerged vegetation type, coverage information, sediment type information, and the corresponding spatial distribution map of the restored water area. The monitoring results are compared and analyzed with the preset restoration goals: vegetation coverage growth rate, species composition changes, and spatial distribution pattern evolution. Subsequent restoration strategies or monitoring plans are adjusted based on the analysis results: for areas with low recovery efficiency, the restoration plan is adjusted; for areas with high efficiency, it is continued; and based on the monitoring data, the model parameters and evaluation system are optimized.

[0150] The following two embodiments briefly illustrate the implementation process of the present invention.

[0151] Example 1

[0152] Use an underwater camera to take a group of underwater photos of the lake, and use simple underwater photo color extraction to identify whether there are submerged plants at the point. Through the supervised classification method, the type of submerged plants is identified. Through color extraction and unsupervised classification of multiple groups of underwater photos, attribute tables of different points are obtained. The attribute table includes whether there are submerged vegetation and what kind of submerged vegetation. Through spatial interpolation of the attribute table, a preliminary spatial distribution map of whether submerged plants are distributed at different points and what kind of plants are distributed is obtained. Combined with manual visual underwater photos, underwater photos of different submerged vegetation are screened, and further supervised classification is performed to improve the classification results and update the attribute table. The specific underwater characteristics of the water body are obtained through spatial interpolation. The recognition results in this embodiment are as follows. Figure 3 As shown, Figure 3 On the left is a picture of underwater submerged vegetation. Figure 3The right side shows the submerged plants identified by supervised classification. This embodiment shows the situation in which the present invention adaptively selects unsupervised classification processing in step S2.

[0153] Example 2

[0154] Training samples were prepared for the supervised classification algorithm. Sediments with varying submerged vegetation and pollutant content were collected. Spectral analysis and image acquisition were performed in the laboratory on samples with varying sediment pollutant content to extract characteristic parameters, including sediment color, texture, and infrared spectral characteristics.

[0155] A series of underwater photos of the bottom mud are obtained by sailing an unmanned boat. The pictures and features extracted in the previous step are used as training samples. The series of underwater photos are classified through training samples and supervision. After the classification is completed, a monitoring point attribute table containing this information is generated, and the relevant attribute data of each monitoring point is recorded. Spatial interpolation is performed to obtain an intuitive spatial distribution map of the bottom mud pollution degree. According to the pollution degree combined with the water depth distribution and vegetation distribution characteristics, a priority time sequence plan for submerged vegetation restoration is formulated. The sample recognition results in this embodiment are as follows. Figure 4 As shown, Figure 4 The first row represents the sediment samples, and the second row represents the pollution degree identified by supervised classification. Figure 4 From left to right, the sediment with low pollution content (left), the sediment with medium pollution content (center), and the sediment with heavy pollution content (right) are respectively represented. This embodiment shows the scenario of applying the supervised classification model in step S2 of the present invention.

[0156] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas, characterized in that: include: S1. Use underwater image acquisition equipment to collect underwater image sequences in the target waters along a preset route, and simultaneously record the geographic coordinates and water depth data corresponding to the underwater images to form an underwater image dataset with geographic coordinates and water depth information; S2. Use an image classification model to process and classify the underwater image dataset to identify the submerged vegetation type, coverage information, and sediment type information at each monitoring point. Manually verify the low-confidence classification results using a composite confidence level, and use the verification results as feedback to optimize the model parameters of the image classification model. S3. Using a spatial interpolation algorithm, using the submerged vegetation type, coverage information, sediment type information, and water depth data of each monitoring point, perform spatial interpolation processing to generate a spatial distribution map of the target water area; S4. Utilize the spatial distribution map of the target waters, combined with preset ecological evaluation factors, and conduct a quantitative assessment through the comprehensive ecological suitability evaluation method to determine the types of submerged vegetation restoration in the target waters, suitable restoration areas, and corresponding restoration priorities.

2. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 1, characterized in that: In step S1, the underwater image acquisition equipment includes a high-definition underwater camera, a depth sounder and a GPS locator installed on a cruise boat; the preset route is a spiral or zigzag route; the underwater image acquisition equipment moves along the route and collects data.

3. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 1, characterized in that: In step S2, the image classification model is an adaptive multi-threshold underwater image enhancement and classification integrated model, which sequentially performs a series of image processing operations on underwater images, including image enhancement preprocessing, multi-scale boundary enhancement processing, and adaptive classification processing, to obtain preliminary classification labels for submerged vegetation at each monitoring point and preliminary classification labels for bottom sediment.

4. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 3, characterized in that: The composite confidence is used to evaluate the preliminary classification labels and is calculated as follows: Conf(p)=γ1·P max (p)+γ2·(1-Entropy(p))+γ3·S spatial (p) Among them, Conf(p) is the classification confidence score of point p; P max (p) is the highest category probability value in the classification result corresponding to point p; Entropy(p) is the normalized probability distribution entropy of the classification result corresponding to point p; S spatial (p) is the spatial consistency score between the classification result of point p and the classification result of its neighborhood; γ1, γ2, γ3 are preset weight coefficients, satisfying γ1+γ2+γ3=1; If the Conf(p) of point p is lower than the preset confidence threshold Conf threshold , then the preliminary classification label of the point is marked as a label to be manually verified.

5. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 1, characterized in that: In step S3, the spatial interpolation algorithm integrates the hydrodynamically corrected spatial correlation structure model and the ecological threshold constraint conditions to perform spatial interpolation processing on the submerged vegetation type, coverage information, sediment type information and water depth data of each monitoring point, and combines the characteristic boundary maintenance interpolation adjustment to generate the spatial distribution map of the target water area, including the submerged vegetation species distribution map, coverage spatial distribution map, sediment type spatial distribution map and water depth spatial distribution map.

6. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 5, characterized in that: The hydrodynamically corrected spatial correlation structure model is based on the preset hydrodynamic correction function After correcting the semivariogram function γ0(h), the relationship is: Where γ(h) is the modified semivariogram function; h is the distance vector between two points in space; is the dominant water flow direction vector; θ is the water flow intensity parameter.

7. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 5, characterized in that: The spatial distribution map generated in step S3 also includes: a spatial distribution map of vegetation species composition and a distribution map of sediment health status.

8. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 1, characterized in that: In step S4, the comprehensive evaluation method of ecological suitability is to use the integrated optimized ecological suitability comprehensive evaluation index IECSI for calculation. The calculation formula of IECSI is: Among them, IECSI(p) is the comprehensive ecological suitability index of point p in the target water area; is the normalized score of the i-th core restriction factor at point p; is the preset weight index of the i-th core limiting factor, satisfying K is the number of core limiting factors; C neighbor (p) is the connectivity index of the neighboring vegetation at point p; α is the neighboring gain adjustment coefficient; is the standardized score of the jth auxiliary support factor at point p; is the preset weight coefficient of the jth auxiliary support factor, satisfying M is the number of auxiliary support factors; Ψ(p) is the timing stability correction coefficient at point p.

9. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 8, characterized in that: The core limiting factors include historical vegetation distribution, water depth suitability and sediment health; the auxiliary supporting factors include sediment particle size suitability and sediment moisture content suitability.

10. The method for monitoring submerged vegetation coverage and rapidly identifying suitable restoration areas according to claim 1, characterized in that: The method further comprises: Based on the comprehensive evaluation index value of ecological suitability at each point, the target water area is divided into priority restoration areas, suboptimal restoration areas, areas to be observed, and unsuitable areas according to the preset index threshold range. The submerged vegetation restoration types and suitable restoration area classification maps of the target water area are output; In areas where submerged vegetation restoration measures have been implemented, steps S1, S2, and S3 are repeated periodically to obtain the submerged vegetation type, coverage information, sediment type information, and corresponding spatial distribution map of the restored water area. The monitoring results are compared and analyzed with the preset restoration targets, and subsequent restoration strategies or monitoring plans are adjusted based on the analysis results.

Images