Lake and reservoir chlorophyll concentration temporal and spatial change analysis method based on unmanned aerial vehicle remote sensing

Abstract

The invention discloses a lake and reservoir chlorophyll concentration temporal and spatial change analysis method based on unmanned aerial vehicle remote sensing, which comprises the following steps: acquiring environmental data of a to-be-analyzed water area, determining a water body sampling point and acquiring chlorophyll test concentration, determining an unmanned aerial vehicle flight path and acquiring a water area spectral image; performing image calibration and image extraction on the water area spectral image to obtain water area spectral data, performing data inversion to obtain chlorophyll predicted concentration, determining climate impact factors and ecological impact factors according to the environmental data, constructing a water area chlorophyll concentration spatio-temporal information model, and performing water area chlorophyll concentration prediction. Inputting environmental data and spectral data of a to-be-analyzed water area into the water area chlorophyll concentration spatio-temporal information model to obtain water area chlorophyll concentration spatio-temporal information, and drawing a water area chlorophyll concentration spatio-temporal contrast diagram. The method not only can improve the efficiency of unmanned aerial vehicle spectrum image acquisition, but also can improve the accuracy of lake and reservoir chlorophyll concentration temporal and spatial change analysis, and is of great significance to lake and reservoir water quality monitoring.

Description

Technical Field

[0001] The present invention relates to the field of data analysis, and particularly to a method for analyzing the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs based on unmanned aerial vehicle (UAV) remote sensing. Background Art

[0002] With the intensification of global climate change and human activities, water body ecosystems such as lakes and reservoirs are facing many challenges. As a key indicator for measuring the degree of water eutrophication and the biomass of phytoplankton, accurate and efficient monitoring of its spatio-temporal variation is of great significance for water quality management and ecological protection.

[0003] Traditional methods for monitoring chlorophyll concentration mainly include manual sampling analysis and satellite remote sensing monitoring. Firstly, the number of sampling points in manual sampling analysis is limited, which is not only time-consuming and laborious, but also difficult to achieve dynamic monitoring of the chlorophyll concentration distribution characteristics of large water areas. In addition, the spatial resolution of satellite remote sensing monitoring is relatively low, and it is greatly affected by environmental and climate fluctuations, making it difficult to ensure the accuracy, timeliness, and continuity of data acquisition. UAVs, with their advantages of strong mobility, convenient operation, and high spatial resolution, can perform refined spectral image acquisition in lakes and reservoirs, providing a new solution for monitoring the chlorophyll concentration in lakes and reservoirs. By combining the specific lake and reservoir environment to determine the UAV flight path and sampling points, efficiently processing the spectral images collected by the UAV, establishing a more accurate and targeted spatio-temporal information model of water area chlorophyll concentration, and designing a method for analyzing the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing, the deficiencies of existing chlorophyll concentration monitoring methods are overcome, providing strong technical support for water quality management, ecological protection, and sustainable development of lakes and reservoirs. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for analyzing the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing.

[0005] To achieve the above object, the present invention is implemented according to the following technical solution:

[0006] The present invention includes the following steps:

[0007] Obtain the environmental data of the water area to be analyzed, determine the water body sampling points according to the environmental data, and perform water body sampling to obtain the measured chlorophyll concentration; the environmental data includes topographic environment, climatic environment, and ecological environment;

[0008] Determine the UAV flight path according to the environmental data and the water body sampling points, collect the water area spectral images based on the UAV flight path, and perform image calibration and image extraction on the water area spectral images to obtain water area spectral data;

[0009] Invert the spectral data of the water area to obtain the predicted chlorophyll concentration, and determine the climate impact factor and ecological impact factor according to the environmental data;

[0010] Construct a spatio-temporal information model of water area chlorophyll concentration, and input the environmental data and spectral data of the water area to be analyzed into the spatio-temporal information model of water area chlorophyll concentration to obtain the spatio-temporal information of water area chlorophyll concentration;

[0011] Draw a spatio-temporal comparison chart of water area chlorophyll concentration according to the spatio-temporal information of water area chlorophyll concentration.

[0012] Furthermore, the method for determining the water body sampling points includes:

[0013] Obtain the topographic environment. The specific steps include: obtain the water area GIS map, and perform image extraction on the water area GIS map to obtain the topographic environment and water body state within the water area;

[0014] Use the grid method to divide the water area into several grids according to the size and shape of the water area, select water body sampling points from the grid vertices, and set the constraint conditions for the sampling points according to the topographic environment and water body state. The expression is:

[0015]

[0016] where β1 is the overall sampling point density threshold, S is the water area area, S i is the sampling point area, is the sampling point geomorphic feature category, is the sampling point water body category, n is the number of sampling points, β2 is the overall sampling point comprehensive score threshold, is the sampling point geomorphic feature score, is the sampling point water body category score, is the sampling point water flow direction score, v i is the sampling point water flow velocity, n1 is the number of near-shore sampling points, n2 is the number of far-shore sampling points, d ni is the off-shore distance of the near-shore sampling point, d fi is the off-shore distance of the far-shore sampling point, α1 is the off-shore distance coefficient, n3 is the number of shallow water area sampling points, h si is the water depth of the shallow water area sampling point, n4 is the number of deep water area sampling points, h di is the water depth of the deep water area sampling point, α2 is the water depth coefficient, and n1 + n2 = n3 + n4 = n;

[0017] Use the geographically weighted regression method to perform data analysis on the sampling points to determine the water area sampling points. The expression is:

[0018]

[0019] where Q iis the trade-off value of sampling point i, and δ0(x i , y i ) is the intercept term of the regression model at the spatial position (x i , y i ) of sampling point i. δ k (x i , y i ) is the regression model coefficient of the k-th independent variable of sampling point i. Value i is the value of the independent variable at the spatial position (x i , y i ) of sampling point i. K is the number of independent variables. ε i is the disturbance term, γ is the spatial distance coefficient, and d ij is the distance between sampling point i and surrounding sampling point j.

[0020] Furthermore, the method for determining the flight path of the drone includes:

[0021] Based on known environmental data and water sampling points, the bitterling optimization algorithm is used to determine the flight path of the drone. Define as the path point, that is, the spawning point of the bitterling individual. D is the dimension of the path point. The bitterling population F = [F1, F2, …, F n T is composed of multiple bitterling individuals, where n is the population size. At the same time, initialize the population to obtain where the boundary is [u, l], and r ∈ (0, 1) is a random perturbation number;

[0022] Search for suitable oysters and determine their occupancy status, that is, search for the status of suitable path points and update the bitterling position. When an oyster is captured by a bitterling, the bitterling position is:

[0023]

[0024]

[0025] where F i t+1 is the updated position of the i-th fish in the (t + 1)-th iteration, is the current position of the i-th fish in the t-th iteration, is the historical position of the i-th fish in the (t - 1)-th iteration. F + is one of the most valuable oysters randomly selected from the population. F * is the best oyster, that is, the optimal path point. λ is a random number in (0, 1). J(t + 1) is the number of steps for the bitterling to move to the escaping oyster at the (t + 1)-th iteration. J(1) is the number of steps for the bitterling to move to the escaping oyster in the initial iteration. t max ​is the maximum number of iterations, U(t) is a random function, a is the restoration power, rand is a perturbation random number;

[0026] When the oyster successfully escapes, the position of the oyster not caught by the bitter fish is:

[0027]

[0028] The male bitter fish that successfully catches the oyster attracts the female bitter fish to produce new bitter fish, and the corresponding bitter fish position is:

[0029]

[0030] where is the position of the newly generated bitter fish at the (t + 1)-th iteration, and R is the distribution radius of the newly generated bitter fish inside the oyster shell;

[0031] The oyster hunts the newly born young bitter fish, and the probability of the lost bitter fish is calculated as:

[0032]

[0033] The objective function f(·) is used to calculate the fitness of the bitter fish that successfully catches the oyster and completes the production and hunting. The population is initialized again, the positions of the bitter fish are updated, and the fitness of the bitter fish is calculated.

[0034] Iterate continuously until the maximum number of iterations is reached. The position of the bitter fish that finally successfully catches the oyster and spawns is the optimal path point.

[0035] Furthermore, the method for obtaining the water area spectral data includes:

[0036] Collect water area spectral images based on the UAV flight path, label them according to the collection time, and perform image calibration on the collected water area spectral images. The specific steps include: radiometric calibration, sensitivity calibration, atmospheric calibration, geometric calibration, image mosaicking, and image enhancement;

[0037] Radiometric calibration converts the DN value into radiance with actual physical meaning through relative radiometric calibration and absolute radiometric calibration operations to eliminate radiometric errors;

[0038] Sensitivity calibration adjusts the brightness and contrast of the image to accurately reflect the lighting conditions of the scene. The expression is:

[0039]

[0040] where ρ nir is the image reflectance, NIR camara is the image signal value, pCam nir is the image gain correction parameter, is the light intensity signal value, li is the optical band category, Y nir is the gray board reflectivity, X nir is the gray board signal value, C atm is the atmospheric correction factor, NLC is the non - linear correction term, I NIR is the image DN value, I blacklevel is the image black - level value, NIR gain is the exposure gain, NIR ctime is the exposure compensation is the distance from the pixel (x, y) to the image center (x * , y * ), k[·] is the compensation coefficient;

[0041] Atmospheric calibration uses an atmospheric transmission model to simulate the influence of the atmosphere on electromagnetic waves to correct the image, so as to eliminate the influence of atmospheric conditions such as solar altitude angle and viewing angle;

[0042] Geometric calibration accurately corrects the geometric shape of the image through GCPs, eliminating image distortion and stitching errors;

[0043] Image mosaicking performs seam - line and feathering processing on the stitched water - area spectral image according to the matching results of UAV flight path points and water - body collection points, and labels according to the positions of water - body collection points;

[0044] Based on the ENVI image - processing software, select the region of interest of the spectral image after image calibration, and use an automatic detection algorithm to extract detailed image parameters to obtain water - area spectral data.

[0045] Furthermore, the method for determining climate impact factors and ecological impact factors according to the environmental data includes:

[0046] Input the climate environmental data into the climate environmental function to obtain the climate impact factor, and the expression is:

[0047]

[0048] where E climate is the climate impact factor, w1 is the temperature - humidity weight, w2 is the light weight, T1 is the current temperature, T0 is the historical average temperature of the corresponding solar term, H1 is the current humidity, H0 is the historical average humidity of the corresponding solar term, h rain is the average rainfall of the current solar term, R1 is the current light intensity, R0 is the historical average light intensity of the corresponding solar term, S1 is the current light radiation, V w is the current wind speed;

[0049] Input the ecological environmental data into the ecological environmental function to obtain the ecological impact factor, and the expression is:

[0050]

[0051] Among them, E ecology is the ecological impact factor, w3 is the production activity weight, m is the number of production activities, f i is the frequency of the i-th production activity, Act i is the score of the i-th production activity, PH1 is the average value of PH randomly sampled from the current water area, PH0 is the PH value with the strongest inhibitory effect on plant growth, ξ is the water body ecological richness coefficient, B i is the water body ecological richness factor, including water body stability, water area biodiversity, grazing pressure, and land utilization rate around the water area.

[0052] Furthermore, the method for obtaining the spatio-temporal information of the chlorophyll concentration in the water area includes:

[0053] Combining the measured chlorophyll concentration, predicted chlorophyll concentration, climate impact factor, and ecological impact factor to form a comprehensive chlorophyll concentration set, and using random forest to divide the comprehensive chlorophyll concentration set into a training set and a test set according to 7:3;

[0054] Constructing a spatio-temporal information model of chlorophyll concentration, the spatio-temporal information model of chlorophyll concentration includes a chlorophyll inversion model, a long short-term memory network, a convolutional neural network, and a fully connected layer;

[0055] The chlorophyll inversion model inverses the spectral data to be analyzed to obtain the predicted chlorophyll concentration, including the UMOC model, the fluorescence baseline model, and the maximum chlorophyll model; the predicted chlorophyll concentration is obtained by weighted fusion of the three inversion results;

[0056] The long short-term memory network processes the time series data of the comprehensive chlorophyll concentration to capture the time characteristics of the comprehensive chlorophyll concentration data; the convolutional neural network processes the spatial data of the comprehensive chlorophyll concentration to capture the spatial characteristics of the comprehensive chlorophyll concentration data; the fully connected layer integrates the prediction results of the long short-term memory network and the convolutional neural network, and extracts data labels and transports them to the output layer;

[0057] Using the Huber loss to reduce the influence of errors and outliers, using the Adam optimizer to automatically adjust the network learning rate, and using cross-validation to evaluate the model performance and select the best model;

[0058] Inputting the environmental data and spectral data of the water area to be analyzed into the spatio-temporal information model of chlorophyll concentration to obtain the chlorophyll concentration and spatio-temporal characteristic information.

[0059] The beneficial effects of the present invention are:

[0060] The present invention is a method for analyzing the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing. Compared with the prior art, the present invention has the following technical effects:

[0061] Through steps such as determining water body sampling points, flight path planning, image calibration, data inversion, constructing influencing factors, and constructing models, the present invention can improve the comprehensiveness of data collection and the ability of data preprocessing in the analysis of the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs, thereby improving the efficiency of the spatio-temporal variation analysis of chlorophyll concentration in lakes and reservoirs. Optimizing the spatio-temporal variation analysis technology of chlorophyll concentration in lakes and reservoirs can greatly save resources, improve work efficiency, and achieve the spatio-temporal variation analysis of chlorophyll concentration in lakes and reservoirs. It can effectively use unmanned aerial vehicle (UAV) remote sensing technology to accurately monitor and analyze the chlorophyll concentration in lakes and reservoirs, providing important support for water resource management and ecological protection. Description of the Drawings

[0062] Figure 1 It is a flowchart of the steps of the method for analyzing the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing of the present invention. Detailed Embodiment

[0063] The present invention will be further described below through specific embodiments. The illustrative embodiments and explanations of the present invention are used to explain the present invention, but do not limit the present invention.

[0064] The method for analyzing the spatio-temporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing of the present invention includes the following steps:

[0065] As Figure 1 shown, in this embodiment, it includes the following steps:

[0066] Obtain the environmental data of the water area to be analyzed, determine the water body sampling points according to the environmental data, and conduct water body sampling to obtain the chlorophyll test concentration; the environmental data includes topographic environment, climate environment, and ecological environment;

[0067] Determine the UAV flight path according to the environmental data and the water body sampling points, collect the water area spectral image based on the UAV flight path, and perform image calibration and image extraction on the water area spectral image to obtain the water area spectral data;

[0068] Invert the water area spectral data to obtain the chlorophyll predicted concentration, and determine the climate influencing factors and ecological influencing factors according to the environmental data;

[0069] Construct a spatio-temporal information model of water area chlorophyll concentration, and input the environmental data and spectral data of the water area to be analyzed into the spatio-temporal information model of water area chlorophyll concentration to obtain the spatio-temporal information of water area chlorophyll concentration;

[0070] Draw a spatio-temporal comparison chart of water area chlorophyll concentration according to the spatio-temporal information of water area chlorophyll concentration.

[0071] In this embodiment, the method for determining the water body sampling points includes:

[0072] Obtain the terrain environment. The specific steps include: obtaining a water area GIS map, and performing image extraction on the water area GIS map to obtain the terrain environment and water body state within the water area; the terrain environment includes the water area boundary and topographic and geomorphic features; the water body state includes the water body type and water flow condition;

[0073] Use the grid method to divide the water area into several grids according to the size and shape of the water area, select water body sampling points from the grid vertices, and set the constraint conditions for the sampling points according to the terrain environment and water body state. The expression is:

[0074]

[0075] where β1 is the overall sampling point density threshold, S is the water area, S i is the sampling point area, is the sampling point geomorphic feature category, is the sampling point water body category, n is the number of sampling points, β2 is the overall sampling point comprehensive score threshold, is the sampling point geomorphic feature score, is the sampling point water body category score, is the sampling point water flow direction score, v i is the sampling point water flow velocity, n1 is the number of near-shore sampling points, n2 is the number of far-shore sampling points, d ni is the distance of the near-shore sampling point from the shore, d fi is the distance of the far-shore sampling point from the shore, α1 is the off-shore distance coefficient, n3 is the number of shallow-water area sampling points, h si is the water depth of the shallow-water area sampling point, n4 is the number of deep-water area sampling points, h di is the water depth of the deep-water area sampling point, α2 is the water depth coefficient, and n1 + n2 = n3 + n4 = n;

[0076] Use the geographically weighted regression method to perform data analysis on the sampling points to determine the water area sampling points. The expression is:

[0077]

[0078] where Q i is the trade-off value of sampling point i, δ0(x i ,y i ) is the intercept term of the regression model at the spatial position (x i ,y i ) of sampling point i, δ k (x i ,y i ) is the regression model coefficient of the kth independent variable of sampling point i, Value i is the spatial position (x i ,y iThe independent variable value at, K is the number of independent variables, ε i is the disturbance term, γ is the spatial distance coefficient, d ij is the distance between sampling point i and surrounding sampling point j.

[0079] In this embodiment, the method for determining the flight path of the drone includes:

[0080] Based on the known environmental data and water body sampling points, the bitter fish optimization algorithm is used to determine the flight path of the drone. Define as the path point, that is, the spawning point of the bitter fish individual, D is the dimension of the path point, and the bitter fish population F = [F1, F2, …, F n T , n is the population quantity. At the same time, initialize the population to obtain where the boundary is [u, l], r ∈ (0, 1) is a random perturbation number;

[0081] Search for suitable oysters and determine their occupancy status, that is, search for the status of suitable path points, and update the bitter fish position. When the oyster is captured by the bitter fish, the bitter fish position is:

[0082]

[0083] where is the updated position of the i-th fish in the (t + 1)-th iteration, is the current position of the i-th fish in the t-th iteration, is the historical position of the i-th fish in the (t - 1)-th iteration, F + is one of the most valuable oysters randomly selected from the population, F * is the best oyster, that is, the optimal path point, λ is a random number in (0, 1), J(t + 1) is the number of steps for the bitter fish to move to the escaping oyster at the (t + 1)-th iteration, J(1) is the number of steps for the bitter fish to move to the escaping oyster in the initial iteration, t max is the maximum number of iterations, U(t) is a random function, a is the reduction power, and rand is a perturbation random number;

[0084] When the oyster successfully escapes, the position of the oyster not caught by the bitter fish is:

[0085]

[0086] The male bitter fish that successfully captures the oyster attracts the female bitter fish to produce new bitter fish, and the corresponding bitter fish position is:

[0087]

[0088] where ​is the position of the newly generated bitter fish at the (t + 1)-th iteration, and R is the distribution radius of the newly generated bitter fish inside the oyster shell;

[0089] The oyster hunts the newly generated young bitter fish, and the probability of the lost bitter fish is calculated as:

[0090]

[0091] The objective function f(·) is used to calculate the fitness of the oysters successfully captured and the bitter fish that have completed production and been hunted. The population is initialized again, the positions of the bitter fish are updated, and the fitness of the bitter fish is calculated.

[0092] Iterate continuously until the maximum number of iterations is reached. The position of the bitter fish that finally successfully captures the oyster and lays eggs is the optimal path point.

[0093] In this embodiment, the method for obtaining the water area spectral data includes:

[0094] Collect water area spectral images based on the flight path of the drone, label them according to the collection time, and perform image calibration on the collected water area spectral images. The specific steps include: radiometric calibration, sensitivity calibration, atmospheric calibration, geometric calibration, image mosaicking, and image enhancement;

[0095] Radiometric calibration converts the DN value into radiance with actual physical meaning through relative radiometric calibration and absolute radiometric calibration operations to eliminate radiometric errors;

[0096] Sensitivity calibration adjusts the brightness and contrast of the image to accurately reflect the lighting conditions of the scene. The expression is:

[0097]

[0098] where ρ nir is the image reflectance, NIR camara is the image signal value, pCam nir is the image gain correction parameter, is the light intensity signal value, l i is the light band category, Y nir is the gray board reflectance, X nir is the gray board signal value, C atm is the atmospheric correction factor, NLC is the non-linear correction term, I NIR is the image DN value, I blacklevel is the image black level value, NIR gain is the exposure gain, NIR ctime is the exposure compensation, is the distance from the pixel (x, y) to the center of the image (x * , y * ), and k[·] is the compensation coefficient;

[0099] Atmospheric calibration uses an atmospheric transmission model to simulate the impact of the atmosphere on electromagnetic waves to correct the image, eliminating the influence of atmospheric conditions such as solar elevation angle and observation angle;

[0100] Geometric calibration precisely corrects the geometric shape of the image through GCPs, eliminating image distortion and stitching errors;

[0101] Image mosaicking performs seam line and feathering processing on the stitched water area spectral image according to the matching results of the UAV flight path points and water body collection points, and labels according to the positions of the water body collection points;

[0102] Based on the ENVI image processing software, select the region of interest of the spectral image after image calibration, and use an automatic detection algorithm to extract detailed image parameters to obtain water area spectral data;

[0103] In the actual evaluation, a UAV is used to collect spectral images of a certain reservoir. After image processing of the spectral images, spectral data is obtained (reflectance corresponding to the positions of 5 water body sampling points, red light band 670nm / green light band 550nm / blue light band 470nm / near-infrared light band 740nm / short-wave infrared light band 1300nm): 1. (145,320,8), 0.34 / absorption valley, 0.56, 0.46, 0.91 / reflectance peak, 0.78; 2. (258,-150,12), 0.29, 0.51, 0.41, 0.87, 0.7; 3. (-180,450,18), 0.24, 0.49, 0.49, 0.82, 0.74; 4. (370,210,22), 0.19, 0.47, 0.32, 0.78, 0.72; 5. (-250,-320,28), 0.14, 0.42, 0.28, 0.74, 0.70.

[0104] In this embodiment, the method for determining the climate impact factor and the ecological impact factor according to the environmental data includes:

[0105] Input the climate environmental data into the climate environmental function to obtain the climate impact factor, and the expression is:

[0106]

[0107] Where E climate is the climate impact factor, w1 is the temperature and humidity weight, w2 is the light weight, T1 is the current temperature, T0 is the historical average temperature corresponding to the solar term, H1 is the current humidity, H0 is the historical average humidity corresponding to the solar term, h rain is the average rainfall of the current solar term, R1 is the current light intensity, R0 is the historical average light intensity corresponding to the solar term, S1 is the current light radiation, V w is the current wind speed;

[0108] Input the ecological environment data into the ecological environment function to obtain the ecological impact factor. The expression is as follows:

[0109]

[0110] Among them, E ecology is the ecological impact factor, w3 is the production activity weight, m is the number of production activities, f i is the frequency of the i-th production activity, Act i is the score of the i-th production activity, PH1 is the average PH value of random sampling in the current water area, PH0 is the PH value with the strongest inhibitory effect on plant growth, ξ is the water body ecological richness coefficient, B i is the water body ecological richness factor, including water body stability, water area biodiversity, grazing pressure and land utilization rate around the water area;

[0111] In the actual evaluation, the historical environmental data of a certain reservoir during the Beginning of Summer period are obtained: average temperature 22°C, average humidity 60%, average rainfall 50mm, average PH value 7.8;

[0112] The environmental data (current temperature, current humidity, current light intensity, current light radiation, current wind speed) of a certain point at 5 monitoring points during the Beginning of Summer period are as follows: 1. 23°C, 62%, 9h / day, 1200W / ㎡, 5m / s; 2. 21°C, 58%, 8.5h / day, 1150W / ㎡, 4m / s; 3. 24°C, 65%, 7.5h / day, 1100W / ㎡, 6m / s; 4. 20°C, 55%, 8h / day, 1130W / ㎡, 3m / s; 5. 25°C, 70%, 9.5h / day, 1250W / ㎡, 7m / s;

[0113] Take w1 = w2 = 0.5. According to the above data, the climate impact factors of the 5 monitoring points are calculated to be 1.3549, 0.4209, 0.4501, 1.3007, and 0.5123 respectively;

[0114] The ecological environment data of a certain reservoir are obtained: the average PH value of random sampling in the current water area is 7.5, the PH value with the strongest inhibitory effect on plant growth is 6.5, the water body stability is 0.8, the water area biodiversity is 0.9, the grazing pressure is 0.7, the land utilization rate around the water area is 60%, the production activity data are (0.6, 80), (0.7, 90), (0.5, 70), w3 = 0.5, and the water body ecological richness coefficient is 1.2; the ecological impact factor calculated according to the ecological environment function is 2.6854.

[0115] In this embodiment, the method for obtaining the spatio-temporal information of the chlorophyll concentration in the water area includes:

[0116] The chlorophyll test concentration, chlorophyll predicted concentration, climate impact factors, and ecological impact factors are combined to form a comprehensive chlorophyll concentration set. The comprehensive chlorophyll concentration set is divided into a training set and a test set in a ratio of 7:3 using random forest.

[0117] A spatio-temporal information model of chlorophyll concentration is constructed. The spatio-temporal information model of chlorophyll concentration includes a chlorophyll inversion model, a long short-term memory network, a convolutional neural network, and a fully connected layer.

[0118] The chlorophyll inversion model inverses the spectral data to be analyzed to obtain the chlorophyll predicted concentration, including the UMOC model, the fluorescence baseline model, and the maximum chlorophyll model. The chlorophyll predicted concentration is obtained by weighted fusion of the three inversion results.

[0119] The long short-term memory network processes the time series data of the comprehensive chlorophyll concentration to capture the time characteristics of the comprehensive chlorophyll concentration data. The convolutional neural network processes the spatial data of the comprehensive chlorophyll concentration to capture the spatial characteristics of the comprehensive chlorophyll concentration data. The fully connected layer integrates the prediction results of the long short-term memory network and the convolutional neural network, and extracts the data labels and sends them to the output layer.

[0120] The Huber loss is used to reduce the influence of errors and outliers. The Adam optimizer is used to automatically adjust the network learning rate. Cross-validation is used to evaluate the model performance and select the best model.

[0121] The environmental data and spectral data of the water area to be analyzed are input into the spatio-temporal information model of chlorophyll concentration to obtain the chlorophyll concentration and spatio-temporal characteristic information.

[0122] In the actual evaluation, the spectral data is inversed using the inversion model, and the inversion results are weighted and fused to obtain the chlorophyll predicted concentrations at 5 sampling points: 3.5 μg / L, 7.0 μg / L, 10.5 μg / L, 14.0 μg / L, 18.5 μg / L. The chlorophyll predicted concentrations at the 5 sampling points and the environmental data are input into the trained spatio-temporal information model of chlorophyll concentration to obtain the chlorophyll concentration and location information at a certain time point during the Beginning of Summer period: 1. (145, 320, 8), 3.0 μg / L; 2. (258, -150, 12), 6.5 μg / L; 3. (-180, 450, 18), 9.8 μg / L; 4. (370, 210, 22), 13.2 μg / L; 5. (-250, -320, 28), 17.0 μg / L.

[0123] Taking the water body type of the water area as the base map, the chlorophyll concentration data at the same time is determined according to the time label. The chlorophyll concentration deviation at different positions is calculated according to the standard chlorophyll concentration, and the chlorophyll concentration spatial distribution map is drawn according to the base map.

[0124] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for analyzing the spatiotemporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing, characterized in that: The following steps are involved: S1. Obtain environmental data of the water area to be analyzed, determine water sampling points according to the environmental data, and perform water sampling to obtain chlorophyll test concentration; the environmental data includes topographic environment, climate environment and ecological environment; S2. Determine a flight path of the drone according to the environmental data and the water body sampling points, collect a water area spectral image based on the drone flight path, and perform image calibration and image extraction on the water area spectral image to obtain water area spectral data; S3, inverting the water area spectral data to obtain predicted chlorophyll concentration, and determining climate impact factors and ecological impact factors according to the environmental data; S4, constructing a spatiotemporal information model of chlorophyll concentration in water areas, and inputting environmental data and spectral images of the water areas to be analyzed into the spatiotemporal information model of chlorophyll concentration in water areas to obtain spatiotemporal information of chlorophyll concentration in water areas; S5. Draw a spatiotemporal comparison diagram of chlorophyll concentration in the water area according to the spatiotemporal information of chlorophyll concentration in the water area.

2. The method for analyzing the spatiotemporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing according to claim 1 is characterized in that: The method for determining the water body sampling point comprises: Acquiring the terrain environment, the specific steps include: acquiring a water area GIS map, performing image extraction on the water area GIS map to obtain the terrain environment and water body status within the water area; The grid method is used to divide the water area into several grids according to the size and shape of the water area. Water sampling points are selected from the grid vertices. The constraints of the sampling points are set according to the terrain environment and water state. The expression is: Where β1 is the overall sampling point density threshold, S is the water area, S i is the sampling point area, C gi is the geomorphic feature category of the sampling point, C wi is the water body category of the sampling point, n is the number of sampling points, β2 is the overall sampling point comprehensive score threshold, P gi Score the geomorphic characteristics of the sampling points, P wi Score the water category of the sampling point, C vi Score the water flow direction at the sampling point, v i is the water velocity at the sampling point, n1 is the number of nearshore sampling points, n2 is the number of farshore sampling points, d ni is the distance from the shore sampling point to the shore, d fi is the distance from the far-shore sampling point to the shore, α1 is the offshore distance coefficient, n3 is the number of sampling points in the shallow water area, h si is the depth of the sampling point in the shallow water area, n4 is the number of sampling points in the deep water area, h di is the water depth of the sampling point in the deep water area, α2 is the water depth coefficient, and n1+n2=n3+n4=n; The geographically weighted regression method was used to analyze the data of the sampling points and determine the water sampling points. The expression is: Where Q i is the trade-off value of sampling point i, δ0(x i ,y i ) is the spatial position of sampling point i (x i ,y i ) is the intercept term of the regression model, δ k (x i ,y i ) is the regression model coefficient of the kth independent variable at sampling point i, Value i is the spatial position of sampling point i (x i ,y i ), K is the number of independent variables, ε i is the disturbance term, γ is the spatial distance coefficient, d ij is the distance between sampling point i and surrounding sampling points j.

3. The method for analyzing the spatiotemporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing according to claim 1 is characterized in that: The method for determining the flight path of the drone comprises: Based on the known environmental data and water sampling points, the flight path of the UAV is determined by using the bitter fish optimization algorithm, and F is defined. i =[F i 1 ,F i 2 ,…,F i D ] is the path point, i.e. the spawning point of the bitter fish individual, D is the dimension of the path point, and the bitter fish population F is composed of multiple bitter fish individuals. n ] T , n is the population size, and the population is initialized to obtain The boundary is [u,l], r∈(0,1) is the random perturbation number; Search for a suitable oyster and determine its occupancy status, that is, search for a suitable path point status, update the position of the bitter fish, and when the oyster is captured by the bitter fish, the position of the bitter fish is: in is the updated position of the ith fish in the t+1th iteration, is the current position of the ith fish in the tth iteration, is the historical position of the ith fish in the t-1th iteration, F + For one of the most worthwhile oysters in the randomly selected population, F * is the best oyster, i.e. the optimal path point, λ is a random number between (0,1), J(t+1) is the number of steps taken by the bitter fish to move to the escaped oyster in the t+1th iteration, J(1) is the number of steps taken by the bitter fish to move to the escaped oyster in the initial iteration, t max is the maximum number of iterations, U(t) is a random function, a is the restoration power, and rand is the perturbation random number; When the oysters successfully escape, the positions of the oysters not caught by the bittern are: The male bittern that successfully captures oysters attracts female bitterns to produce new bitterns. The corresponding bittern positions are: in is the position of the newly generated bitter fish at the t+1th iteration, and R is the distribution radius of the newly generated bitter fish inside the oyster shell; Oysters hunt newly produced young bitterlings. The probability of losing bitterlings is calculated as: The objective function f(·) is used to calculate the fitness of bittern that successfully captures oysters and completes production and is hunted, and the population is initialized again, the position of bittern is updated, and the fitness of bittern is calculated. Iterate continuously until the maximum number of iterations is reached, and the position of the bittern that successfully captures the oysters and lays eggs is the optimal path point.

4. The method for analyzing the spatiotemporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing according to claim 1 is characterized in that: The method for obtaining the water area spectral data comprises: Based on the flight path of the UAV, a water spectral image is collected, a label is attached according to the collection time, and the collected water spectral image is calibrated. The specific steps include: radiation calibration, sensitivity calibration, atmosphere calibration, geometric calibration, image mosaicking and image enhancement; Radiation calibration converts DN value into radiation brightness with actual physical meaning through relative radiation calibration and absolute radiation calibration to eliminate radiation error. Sensitivity calibration adjusts the brightness and contrast of the image to accurately reflect the lighting conditions of the scene. The expression is: where ρ nir is the image reflectance, NIR camara is the image signal value, pCam nir is the image gain correction parameter, NIR li ·pl inir is the light intensity signal value, l i is the optical band category, Y nir is the reflectivity of the gray plate, X nir is the gray plate signal value, C atm is the atmospheric correction factor, NLC is the nonlinear correction term, I NIR is the image DN value, I blacklevel is the image black level value, NIR gain is the exposure gain, NIR ctime For exposure compensation, is the distance from pixel (x, y) to the image center (x * ,y * ) is the distance, k[·] is the compensation coefficient; Atmospheric calibration uses an atmospheric transmission model to simulate the effect of the atmosphere on electromagnetic waves to correct images in order to eliminate the effects of atmospheric conditions such as solar altitude angle and observation angle. Geometric calibration accurately corrects the geometric shape of the image through GCPs to eliminate image distortion and stitching errors; Image mosaicking is to perform edge-joining and feathering processing on the spliced ​​water spectral image according to the matching results of the UAV flight path points and the water body collection points, and label the water body collection points according to their positions; The region of interest of the spectral image after image calibration was selected based on the ENVI image processing software, and the automatic detection algorithm was used to extract detailed image parameters to obtain the water area spectral data.

5. The method for analyzing the spatiotemporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing according to claim 1 is characterized in that: The method for determining climate impact factors and ecological impact factors according to the environmental data comprises: Input the climate and environment data into the climate and environment function to obtain the climate impact factor, the expression is: Where E climate is the climate influencing factor, w1 is the temperature and humidity weight, w2 is the light weight, T1 is the current temperature, T0 is the historical average temperature of the corresponding solar term, H1 is the current humidity, H0 is the historical average humidity of the corresponding solar term, and h rain is the average rainfall of the current solar term, R1 is the current light intensity, R0 is the historical average light intensity of the corresponding solar term, S1 is the current light radiation, V w is the current wind speed; Input the ecological environment data into the ecological environment function to obtain the ecological impact factor, the expression is: Where E ecology is the ecological impact factor, w3 is the production activity weight, m ​​is the number of production activities, f i is the frequency of the i-th production activity, Act i is the score of the i-th production activity, PH1 is the mean pH value of the current water area randomly sampled, PH0 is the pH value with the strongest inhibitory effect on plant growth, ξ is the water body ecological richness coefficient, B i It is a factor of water body ecological richness, including water body stability, water body biodiversity, grazing pressure and land use rate around water bodies.

6. The method for analyzing the spatiotemporal variation of chlorophyll concentration in lakes and reservoirs based on UAV remote sensing according to claim 1 is characterized in that: The method for obtaining the spatiotemporal information of chlorophyll concentration in the water area comprises: The chlorophyll test concentration, chlorophyll predicted concentration, climate influencing factors and ecological influencing factors are combined into a comprehensive chlorophyll concentration set, and the comprehensive chlorophyll concentration set is divided into a training set and a test set in a ratio of 7:3 using random forest. Constructing a chlorophyll concentration spatiotemporal information model, wherein the chlorophyll concentration spatiotemporal information model includes a chlorophyll inversion model, a long short-term memory network, a convolutional neural network, and a fully connected layer; The chlorophyll inversion model inverts the spectral data to be analyzed to obtain the predicted chlorophyll concentration, including the UMOC model, the fluorescence baseline model and the maximum chlorophyll model; the predicted chlorophyll concentration is obtained by weighted fusion of the three inversion results; The long short-term memory network processes the time series data of the comprehensive chlorophyll concentration and captures the time characteristics of the comprehensive chlorophyll concentration data; the convolutional neural network processes the spatial data of the comprehensive chlorophyll concentration and captures the spatial characteristics of the comprehensive chlorophyll concentration data; the fully connected layer integrates the prediction results of the long short-term memory network and the convolutional neural network, and extracts the data labels and transmits them to the output layer; Huber loss is used to reduce the impact of errors and outliers, Adam optimizer is used to automatically adjust the network learning rate, and cross-validation is used to evaluate model performance and select the best model; The environmental data and spectral data of the water area to be analyzed are input into the chlorophyll concentration spatiotemporal information model to obtain the chlorophyll concentration and spatiotemporal characteristic information.

Images