Method for monitoring dissolved oxygen concentration of freshwater aquaculture pond based on multispectral image

Through the multi-spectral imaging technology and spectral index model of the UAV, the problem that traditional methods are difficult to monitor the spatial distribution of dissolved oxygen in freshwater crab farming ponds is solved, real-time and accurate dissolved oxygen monitoring is achieved, and aquaculture efficiency and aquatic product quality are improved.

CN118937252APending Publication Date: 2024-11-12NONGXIN (NANJING) SMART AGRI RES INST CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202410839464.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-06-26
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Traditional methods are difficult to fully grasp the spatial distribution information of dissolved oxygen concentration in freshwater crab farming ponds, and cannot monitor the water quality status of the entire breeding area in real time, quickly and accurately, resulting in low breeding efficiency and high risk.

Method used

Using the dissolved oxygen monitoring method of freshwater aquaculture ponds based on multispectral images of drones, the spectral index model of chlorophyll a concentration and dissolved oxygen concentration was constructed, and the forest algorithm was used to estimate it. and draw spatial distribution maps of dissolved oxygen concentration.

Benefits of technology

It realizes comprehensive, real-time and accurate monitoring of dissolved oxygen concentration in aquaculture ponds, provides basic technologies for intelligent and digital management, and improves aquatic product quality and aquaculture efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118937252A_ABST
    Figure CN118937252A_ABST
Patent Text Reader

Abstract

The invention discloses a freshwater aquaculture pond dissolved oxygen concentration monitoring method based on a multispectral image, which comprises the following steps: a, in a crab aquaculture base, carrying out multispectral image data acquisition on the whole aquaculture pond by an unmanned aerial vehicle, synchronously acquiring water sample data for local aquaculture ponds, and obtaining corresponding chlorophyll a concentration and dissolved oxygen concentration; b, extracting final spectral reflectance data corresponding to the water sample data collected by the local aquaculture pond, c, screening a sensitive spectral index of dissolved oxygen concentration and a sensitive spectral index of chlorophyll a concentration by utilizing the final spectral reflectance data of the water sample data collected in the step b, d, constructing a chlorophyll a concentration estimation model and a dissolved oxygen concentration estimation model by using a forest algorithm; e, drawing a chlorophyll a concentration spatial distribution diagram of the whole culture pond; f, drawing a dissolved oxygen concentration spatial distribution diagram of the whole culture pond; according to the invention, the dissolved oxygen concentration of the aquaculture pond can be effectively estimated, and the dissolved oxygen concentration of the aquaculture pond can be comprehensively and accurately monitored in real time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of spectral data processing, and in particular to a method for monitoring dissolved oxygen concentration in a freshwater aquaculture pond based on multispectral imaging. Background Art

[0002] The dissolved oxygen concentration in crab aquaculture water is one of the key factors that determine the success or failure of crab aquaculture. The concentration of dissolved oxygen has an important influence on the physiological characteristics and breeding effects of crabs. The appropriate dissolved oxygen concentration helps to improve the metabolic efficiency of crabs and promote growth and development, while hypoxia may lead to metabolic abnormalities and growth retardation. In addition, the dissolved oxygen concentration is closely related to the behavioral performance and reproductive activities of crabs. Adequate oxygen helps to maintain their normal behavior and improve reproductive effects. The dissolved oxygen concentration in large ponds is easily affected by many factors. Not only that, the spatial heterogeneity of water quality in the water area is also very significant. Traditional sensor monitoring methods cannot accurately control the overall water quality of the entire breeding area. If the number of sensors is increased, the monitoring cost will increase and the difficulty of operation will increase. Therefore, in the crab breeding industry, real-time, rapid and accurate grasp of the dissolved oxygen concentration of the pond will help improve breeding efficiency, promote the healthy growth of crabs, reduce breeding risks, and provide an important basis for the scientific management and sustainable breeding of breeding ponds.

[0003] The UAV remote sensing system provides a convenient means for the analysis and estimation of water quality parameters of small and medium-sized inland water bodies, with its advantages of flexibility, cost-effectiveness, low-altitude flight with little atmospheric interference, simple operation, and high spatial resolution of images. It can formulate targeted monitoring plans based on the characteristics of the monitored waters. Considering that the water quality of aquaculture ponds has significant spatial variability, in order to effectively solve this problem, we fully utilize the advantages of UAV imaging technology, and its ability to quickly and flexibly obtain high-resolution images provides an effective means for the present invention.

[0004] Up to now, most of the monitoring methods for dissolved oxygen are manual sampling and traditional sensor monitoring methods; while a few researchers use neural network methods, spatiotemporal prediction models and other methods to predict dissolved oxygen by taking water quality parameters and meteorological environment as model input features. Although the above methods can achieve good monitoring accuracy, they can only obtain dissolved oxygen concentration information at local points in the water body, and it is difficult to obtain the spatial distribution of dissolved oxygen in the entire aquaculture pond, and cannot solve the problem of spatial heterogeneity of water quality in the entire aquaculture pond.

[0005] Therefore, in order to fully grasp the spatial distribution information of dissolved oxygen concentration in crab breeding ponds, a rapid monitoring method for dissolved oxygen in freshwater breeding ponds based on multispectral imaging of unmanned aerial vehicles was developed. Chlorophyll a concentration and spectral index were used as input features of the dissolved oxygen estimation model to improve the estimation accuracy of dissolved oxygen, and the spatial distribution map of dissolved oxygen concentration was drawn, providing a new technical support for remote sensing technology in the field of dissolved oxygen monitoring. Summary of the invention

[0006] The present invention is mainly aimed at freshwater crab breeding areas. The water depth in the breeding areas is within 1 meter, and the water level in the crab breeding area gradually deepens as the crabs grow. The water depth is continuously changing, and the range of change is 0.3 meters to 1 meter. Since methods such as manual sampling and sensor monitoring can only obtain the dissolved oxygen concentration information of local points in the water body, it is impossible to fully grasp the spatial distribution information of the dissolved oxygen concentration in the breeding pond, so it is necessary to develop a new monitoring method.

[0007] In order to solve the problem of being unable to fully grasp the spatial distribution information of dissolved oxygen concentration in breeding ponds, this patent provides a dissolved oxygen monitoring method for freshwater breeding ponds based on multispectral imaging, which can effectively estimate the dissolved oxygen concentration in breeding ponds and draw the spatial distribution information of dissolved oxygen concentration in the entire breeding pond, thereby realizing comprehensive, real-time and accurate monitoring of dissolved oxygen concentration in breeding ponds, and providing basic technical support for the intelligent and digital management of dissolved oxygen information monitoring in freshwater crab breeding ponds.

[0008] The technical solutions adopted to achieve the above objectives are:

[0009] The method for monitoring dissolved oxygen in freshwater aquaculture ponds based on multispectral imaging includes the following steps:

[0010] a. In the crab breeding base, the drone collects multispectral image data of the entire breeding pond, and simultaneously collects water sample data for local breeding ponds. After the water sample data is collected, it is sent to the laboratory for chemical testing to obtain the corresponding chlorophyll a concentration and dissolved oxygen concentration. The obtained chlorophyll a concentration and dissolved oxygen concentration are divided into 7:3, 70% of the data is used as the model training set data, and 30% of the data is used as the model validation set data;

[0011] b. Based on the multispectral image data of the drone obtained in step a, the final spectral reflectance data corresponding to the water sample data collected from the local aquaculture pond is extracted. The multispectral data has 5 bands of data, namely blue (R1), green (R2), red (R3), red edge (R4), and near infrared (R5). The central wavelengths are [R1 (450nm), R2 (560nm), R3 (650nm), R4 (730nm), R5 (840nm)]; it is necessary to construct a 6*6 matrix region of interest based on the pixel where the spatial position of the field sampling point is located, and take the mean of the spectral reflectance in the region of interest as the final spectral reflectance of the sampling point.

[0012] c. Using the final spectral reflectance data of the water sample data collected in step b, the final spectral reflectance data of each band are combined, and the spectral index of chlorophyll a concentration or dissolved oxygen concentration is calculated by addition, subtraction, multiplication and division or differential derivation. According to the correlation coefficient |r|>0.4, the sensitive spectral index that meets the conditions is screened; the calculation formula of the correlation coefficient r is as follows:

[0013]

[0014] Among them, x i is the measured value of chlorophyll a concentration or dissolved oxygen concentration, y i is the spectral index corresponding to chlorophyll a concentration or dissolved oxygen concentration, and They are x i and i The average value of

[0015] d. Based on the sensitive spectral index of dissolved oxygen concentration and the sensitive spectral index of chlorophyll a concentration screened in step c, a chlorophyll a concentration estimation model and a dissolved oxygen concentration estimation model are constructed using the forest algorithm; the sensitive spectral index of chlorophyll a concentration corresponding to the training set data is used as the input feature of the model, the chlorophyll a concentration is estimated using the validation set data, and then the model performance of the chlorophyll a concentration estimation model is evaluated; the chlorophyll a concentration estimation value and the sensitive spectral index of dissolved oxygen concentration are used as the input features of the dissolved oxygen concentration estimation model, the dissolved oxygen concentration is estimated, and then the model performance of the dissolved oxygen concentration estimation model is evaluated;

[0016] When evaluating, use the coefficient of determination R 2 , RMSE were used to evaluate the performance of the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model. 2 >0.7 and the root mean square error RMSE is lower than 2 μg / L, the chlorophyll a concentration estimation model is considered to have good performance and meet the requirements; when the determination coefficient R 2>0.7 and the root mean square error RMSE is lower than 3 μg / L, the dissolved oxygen concentration estimation model is considered to have good performance and meet the requirements; 2 The calculation formula of RMSE is as follows:

[0017]

[0018] In the formula, k i are the predicted values ​​of water quality parameters chlorophyll a concentration and dissolved oxygen concentration, are the measured values ​​of the water quality parameters chlorophyll a concentration and dissolved oxygen concentration, is the average value of the measured values ​​of water quality parameter concentration and dissolved oxygen concentration, and n is the number of sampling points;

[0019] e. Based on the multispectral image data of the entire breeding pond collected by the drone, repeat steps c and d to calculate the sensitive spectral index of the chlorophyll a concentration of the entire breeding pond; then use the sensitive spectral index of the chlorophyll a concentration of the entire breeding pond as the model input feature, import it into the chlorophyll a concentration estimation model trained in step d, and draw a spatial distribution map of the chlorophyll a concentration of the entire breeding pond;

[0020] f. Based on step c, step d and step e, calculate the sensitive spectral index of the dissolved oxygen concentration of the entire breeding pond, use the sensitive spectral index of the dissolved oxygen concentration of the entire breeding pond and the spatial distribution map of chlorophyll a concentration as model input features, import them into the dissolved oxygen concentration estimation model trained in step d, and draw the spatial distribution map of the dissolved oxygen concentration of the entire breeding pond.

[0021] Furthermore, in step a, the water sample data collected from local breeding ponds and the multispectral image data collected by drones should be collected between 9:00 and 10:00 in the morning or between 15:00 and 16:00 in the afternoon; when collecting water sample data from local breeding ponds, the sampling depth is 30 cm underwater. The water sample data collection method is the five-point mixing method. First, the collected water samples are mixed evenly using an opaque container; then the water samples are placed in 2L and 500ml opaque water bottles respectively, where the 2L water sample is used to detect the chlorophyll a concentration, and the 500ml water sample is used to detect the dissolved oxygen concentration. 2ml of magnesium carbonate suspension is dripped into the 2L sampling water bottle to prevent acidification from causing pigment dissolution; after the collection is completed, the 2L and 500m water bottles are promptly placed in an insulated box with ice bags for low temperature, light-proof storage, and transportation, and sent to the laboratory for inspection within 8 hours; the number of water sample data collected should be no less than 60.

[0022] Furthermore, when extracting the spectral reflectance data corresponding to the water sample data collected from the local aquaculture pond in step b, it is necessary to construct a 6*6 matrix region of interest based on the pixel where the spatial position of the field sampling point is located, and take the mean of the spectral reflectance in the region of interest as the final spectral reflectance of the sampling point.

[0023] Furthermore, in the modeling process in step d, if the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model cannot meet the evaluation criteria, it is necessary to establish more spectral indices and use more correlated sensitive spectral indices as model input features to improve the performance of the estimation model, and check whether there are outliers in the sample data and eliminate the outliers until the evaluation criteria are met.

[0024] The beneficial effects of the present invention are:

[0025] The present invention is a method for monitoring dissolved oxygen concentration in freshwater aquaculture ponds based on multispectral imaging. Based on spectral analysis technology and the relationship between chlorophyll a and dissolved oxygen concentration, the method can effectively estimate the dissolved oxygen concentration in the aquaculture pond and draw the spatial distribution information of the dissolved oxygen concentration in the entire aquaculture pond, which is helpful to grasp the spatial distribution information of the dissolved oxygen concentration in the pond in real time, quickly and accurately, and realize comprehensive, real-time and accurate monitoring of the dissolved oxygen concentration in the aquaculture pond, providing basic technical support for the intelligent and digital management of dissolved oxygen concentration information monitoring in freshwater crab aquaculture ponds, and improving the quality of aquatic products. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the present invention;

[0027] Figure 2 is a table for constructing and calculating the spectral characteristic index in an embodiment of the present invention;

[0028] Figure 3 This is a bar graph of the correlation analysis of chlorophyll a concentration in the present invention;

[0029] Figure 4 It is a bar graph of correlation analysis of dissolved oxygen concentration in the present invention;

[0030] Figure 5 It is a scatter plot of the chlorophyll a concentration estimation model in the present invention;

[0031] Figure 6 It is a scatter plot of the dissolved oxygen concentration estimation model in the present invention;

[0032] Figure 7 This is a spatial distribution diagram of dissolved oxygen concentration in the aquaculture pond in the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described below in conjunction with the accompanying drawings.

[0034] At present, the inversion of photosensitive water quality parameters has gradually matured, but the optical properties of non-photosensitive parameters, such as dissolved oxygen (DO), are weak, and have little effect on the spectrum captured by the spectral sensor. It is difficult to directly obtain it through spectral feature inversion, which is more significant in the water environment of freshwater aquaculture ponds, because there are usually more organic matter and particulate matter in the water, which increases the complexity of the spectral signal. In addition, freshwater aquaculture ponds are rich in aquatic plants such as waterweed, and the chlorophyll a concentration in the water is relatively high. There is a close relationship between the chlorophyll a concentration and the dissolved oxygen concentration in the water. The relationship between the two is mainly reflected in its participation in the photosynthesis process. Chlorophyll a, as the main photosynthetic pigment of plants and algae, converts carbon dioxide and water into organic matter through photosynthesis, releases oxygen, and thus increases the dissolved oxygen concentration in the water. It is the main source of dissolved oxygen in freshwater aquaculture ponds. Based on spectral analysis technology and the relationship between chlorophyll a and dissolved oxygen concentration, an effective means is provided for monitoring the dissolved oxygen concentration in freshwater aquaculture ponds by combining spectral features with chlorophyll a concentration, which provides a theoretical basis for monitoring dissolved oxygen concentration.

[0035] like Figure 1-7 As shown, according to the above principle, the present invention proposes a method for monitoring dissolved oxygen in freshwater aquaculture ponds based on multispectral imaging, which includes the following steps:

[0036] a. In the crab breeding base, the drone collects multispectral image data of the entire breeding pond, and simultaneously collects water sample data for local breeding ponds. After the water sample data is collected, it is sent to the laboratory for chemical testing to obtain the corresponding chlorophyll a concentration and dissolved oxygen concentration. The obtained chlorophyll a concentration and dissolved oxygen concentration are divided into 7:3, 70% of the data is used as the model training set data, and 30% of the data is used as the model validation set data;

[0037] b. Based on the multispectral image data of the drone obtained in step a, the final spectral reflectance data corresponding to the water sample data collected from the local aquaculture pond is extracted. The multispectral data has 5 bands of data, namely blue (R1), green (R2), red (R3), red edge (R4), and near infrared (R5). The central wavelengths are [R1 (450nm), R2 (560nm), R3 (650nm), R4 (730nm), R5 (840nm)];

[0038] c. Using the spectral reflectance data of the water sample data collected in step b, the final spectral reflectance data of each band are combined, and the sensitive spectral index of chlorophyll a concentration or dissolved oxygen concentration is calculated by addition, subtraction, multiplication and division or differential derivation. According to the correlation coefficient |r|>0.4, the sensitive spectral index that meets the conditions is screened; the calculation formula of the correlation coefficient r is as follows:

[0039]

[0040] Among them, x i is the measured value of chlorophyll a concentration or dissolved oxygen concentration, y i is the spectral index corresponding to chlorophyll a concentration or dissolved oxygen concentration, and They are x i and i The average value of

[0041] In the present invention, by analyzing and processing the sensitive spectral index, characteristic information of freshwater aquaculture ponds can be extracted. In the field of remote sensing, these technologies and methods play an important role and help to better understand and monitor the environmental and resource changes on the earth's surface.

[0042] d. Based on the sensitive spectral index of dissolved oxygen concentration and the sensitive spectral index of chlorophyll a concentration screened in step c, a chlorophyll a concentration estimation model and a dissolved oxygen concentration estimation model are constructed using the forest algorithm; the sensitive spectral index of chlorophyll a concentration corresponding to the training set data is used as the input feature of the model, the chlorophyll a concentration is estimated using the validation set data, and then the model performance of the chlorophyll a concentration estimation model is evaluated; the chlorophyll a concentration estimation value and the sensitive spectral index of dissolved oxygen concentration are used as the input features of the dissolved oxygen concentration estimation model, the dissolved oxygen concentration is estimated, and then the model performance of the dissolved oxygen concentration estimation model is evaluated;

[0043] When evaluating, use the coefficient of determination R 2 , RMSE, and R 2 It is a statistical indicator used to evaluate the goodness of fit of a regression model. RMSE is used to measure the deviation between the observed value and the true value. 2 The value range of R is [0,1]. 2 The closer it is to 1, the better the model fit is. 2 The closer it is to 0, the worse the regression line fitting effect is; the range of RMSE is [0, +∞), when the predicted value is completely consistent with the true value, it is equal to 0, that is, a perfect model; the larger the error, the larger the value. In the present invention, when the determination coefficient R 2 >0.7 and the root mean square error RMSE is lower than 2 μg / L, the chlorophyll a concentration estimation model is considered to have good performance and meet the requirements; when the determination coefficient R 2 >0.7 and the root mean square error RMSE is lower than 3 μg / L, the dissolved oxygen concentration estimation model is considered to have good performance and meet the requirements; 2 The calculation formula of RMSE is as follows:

[0044]

[0045] In the formula, k i are the predicted values ​​of water quality parameters chlorophyll a concentration and dissolved oxygen concentration, are the measured values ​​of the water quality parameters chlorophyll a concentration and dissolved oxygen concentration, is the average value of the measured values ​​of water quality parameter concentration and dissolved oxygen concentration, and n is the number of sampling points;

[0046] e. Based on the multispectral image data of the entire breeding pond collected by the drone, repeat steps c and d to calculate the sensitive spectral index of the chlorophyll a concentration of the entire breeding pond; then use the sensitive spectral index of the chlorophyll a concentration of the entire breeding pond as the model input feature, import it into the chlorophyll a concentration estimation model trained in step d, and draw a spatial distribution map of the chlorophyll a concentration of the entire breeding pond;

[0047] f. Based on step c, step d and step e, calculate the sensitive spectral index of the dissolved oxygen concentration of the entire breeding pond, use the sensitive spectral index of the dissolved oxygen concentration of the entire breeding pond and the spatial distribution map of chlorophyll a concentration as model input features, import them into the dissolved oxygen concentration estimation model trained in step d, and draw the spatial distribution map of the dissolved oxygen concentration of the entire breeding pond.

[0048] In order to ensure the synchronous collection of UAV multispectral image data and water sample data, the collection of water sample data in the local breeding pond and the collection of multispectral image data by UAV in step a should be selected between 9:00 and 10:00 in the morning or between 15:00 and 16:00 in the afternoon; the UAV image data acquired during this time period can effectively avoid the influence of flares caused by clouds and sun flicker, and reduce the interference of other external factors on the real spectral information of the water surface, thereby ensuring the authenticity of the image data.

[0049] When collecting water sample data in local aquaculture ponds, the sampling depth is 30 cm underwater. The water sample data collection method is the five-point mixing method. First, use an opaque container to mix the collected water samples evenly; then put the water samples into 2L and 500ml opaque water bottles respectively, of which 2L water samples are used to detect chlorophyll a concentration, and 500ml water samples are used to detect dissolved oxygen concentration. 2ml of magnesium carbonate suspension is dripped into the 2L sampling water bottle to prevent acidification from causing pigment dissolution; after the collection is completed, the 2L and 500m water bottles are promptly placed in an insulated box with ice bags at low temperature, away from light, and transported, and sent to the laboratory for inspection within 8 hours; the number of water sample data collected should be no less than 60.

[0050] In order to extract the final spectral reflectance data, when extracting the final spectral reflectance data corresponding to the water sample data collected from the local breeding pond in step b, it is necessary to construct a 6*6 matrix region of interest based on the pixel where the spatial position of the field sampling point is located as the center, and take the mean of the spectral reflectance in the region of interest as the final spectral reflectance of the sampling point. The multispectral data has 5 bands of data, namely blue (R1), green (R2), red (R3), red edge (R4), and near infrared (R5) bands, and their central wavelengths are [R1 (450nm), R2 (560nm), R3 (650nm), R4 (730nm), R5 (840nm)]. In an embodiment of the present invention, the acquired drone multispectral data and water sample measured value data total 60 groups. The measured values ​​of chlorophyll a concentration and dissolved oxygen concentration obtained in step a and their corresponding spectral reflectance data sets are divided into 7:3, with 70% of the data used as the model training set and 30% of the data used as the model verification set, which are used to construct and verify the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model.

[0051] In the present invention, in the modeling process in step d, if the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model cannot meet the evaluation criteria, it is necessary to establish more sensitive spectral indices, and use sensitive spectral indices with higher correlation as model input features to improve the performance of the estimation model, and check whether there are outliers in the sample data, and eliminate the outliers until the evaluation criteria are met.

[0052] In the present invention, python programming language is used, and then the forest algorithm is used to construct a chlorophyll a concentration estimation model and a dissolved oxygen concentration estimation model. The model constructed by the forest algorithm consists of multiple decision trees, which improves the robustness and accuracy of the model by voting or averaging, can process a large amount of high-dimensional data, and reduces the risk of overfitting by integrating multiple decision trees. It is applicable to various data types and problems. In the present invention, the main parameters for adjusting the algorithm constructed by the forest algorithm are n_estimators and max_depth, where the meaning of n_estimators is: the number of sub-datasets generated by sampling with replacement of the original data set, that is, the number of decision trees. If n_estimators is too small, it is easy to underfit, and if it is too large, the model cannot be significantly improved, so n_estimators selects a moderate value, and the adjustment range is [50,1000]. max_depth means the maximum depth of the decision tree. If it is equal to None, it means that the decision tree will not limit the depth of the subtree when building the optimal model. If the model has a large number of samples and features, it is recommended to limit the maximum depth; if the sample size or features are small, the maximum depth is not limited. The adjustment range is [1,10], and the other parameters use the default parameters. The forest algorithm is an existing algorithm and a common algorithm in this field, so it will not be described here.

[0053] In the present invention, the sensitive spectral index is named in the format of Vi, where i is the number of spectral indices; then the sensitive spectral index is correlated with the chlorophyll a concentration and the dissolved oxygen concentration, and the spectral index is screened according to the correlation coefficient |r|>0.4 to screen out the sensitive spectral index of the chlorophyll a concentration and the dissolved oxygen concentration. R1 to R5 represent the reflectance corresponding to different bands, respectively. In the present invention, Figure 2-4 As shown, through the construction and calculation of the sensitive spectral index, the correlation between chlorophyll a concentration and dissolved oxygen concentration is calculated, and the correlation analysis bar chart of chlorophyll a concentration and dissolved oxygen concentration is obtained and screened. After screening, the sensitive spectral index of chlorophyll a concentration is: V12, V18 and V22; the sensitive spectral index of dissolved oxygen concentration is: V17.

[0054] Then, the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model were constructed by the forest algorithm. Figure 5-6 , by verifying the compliance with the indicators, we finally get the spatial distribution map of dissolved oxygen concentration in the entire aquaculture pond, such as Figure 7 shown.

[0055] This embodiment does not impose any formal restrictions on the shape, material, structure, etc. of the patent of the present invention. Any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the patent of the present invention shall fall within the protection scope of the technical solution of the patent of the present invention.

[0056] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some of the technical features therein by equivalents, but these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring dissolved oxygen concentration in freshwater aquaculture ponds based on multispectral imaging, characterized in that: The steps include: a. In the crab breeding base, the drone collects multispectral image data of the entire breeding pond, and simultaneously collects water sample data for local breeding ponds. After the water sample data is collected, it is sent to the laboratory for chemical testing to obtain the corresponding chlorophyll a concentration and dissolved oxygen concentration. The obtained chlorophyll a concentration and dissolved oxygen concentration are divided into 7:3, 70% of the data is used as the model training set data, and 30% of the data is used as the model validation set data; b. Based on the multispectral image data of the unmanned aerial vehicle obtained in step a, the final spectral reflectance data corresponding to the water sample data collected from the local aquaculture pond is extracted. The multispectral data has 5 bands of data, namely blue (R1), green (R2), red (R3), red edge (R4), and near infrared (R5). The central wavelengths are [R1 (450nm), R2 (560nm), R3 (650nm), R4 (730nm), R5 (840nm)]; based on the pixel where the spatial position of the field sampling point is located as the center, a 6*6 matrix region of interest is constructed, and the mean value of the spectral reflectance in the region of interest is taken as the final spectral reflectance of the sampling point; c. Using the final spectral reflectance data of the water sample data collected in step b, the final spectral reflectance data of each band are combined, and the sensitive spectral index of chlorophyll a concentration or dissolved oxygen concentration is calculated by addition, subtraction, multiplication and division or differential derivation. According to the correlation coefficient |r|>0.4, the sensitive spectral index that meets the conditions is screened; the calculation formula of the correlation coefficient r is as follows: Among them, x i is the measured value of chlorophyll a concentration or dissolved oxygen concentration, y i is the sensitive spectral index corresponding to chlorophyll a concentration or dissolved oxygen concentration, and They are x i and i The average value of d. Based on the sensitive spectral index of dissolved oxygen concentration and the sensitive spectral index of chlorophyll a concentration screened in step c, a chlorophyll a concentration estimation model and a dissolved oxygen concentration estimation model are constructed using the forest algorithm; the sensitive spectral index of chlorophyll a concentration corresponding to the training set data is used as the input feature of the model, the chlorophyll a concentration is estimated using the validation set data, and then the model performance of the chlorophyll a concentration estimation model is evaluated; the chlorophyll a concentration estimation value and the sensitive spectral index of dissolved oxygen concentration are used as the input features of the dissolved oxygen concentration estimation model, the dissolved oxygen concentration is estimated, and then the model performance of the dissolved oxygen concentration estimation model is evaluated; When evaluating, use the coefficient of determination R 2 , RMSE were used to evaluate the performance of the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model. 2 >0.7 and the root mean square error RMSE is lower than 2 μg / L, the chlorophyll a concentration estimation model is considered to have good performance and meet the requirements; when the determination coefficient R 2 >0.7 and the root mean square error RMSE is lower than 3 μg / L, the dissolved oxygen concentration estimation model is considered to have good performance and meet the requirements; 2 The calculation formula of RMSE is as follows: In the formula, k i are the predicted values ​​of water quality parameters chlorophyll a concentration and dissolved oxygen concentration, are the measured values ​​of the water quality parameters chlorophyll a concentration and dissolved oxygen concentration, is the average value of the measured values ​​of water quality parameter concentration and dissolved oxygen concentration, and n is the number of sampling points; e. Based on the multispectral image data of the entire breeding pond collected by the drone, repeat steps c and d to calculate the sensitive spectral index of the chlorophyll a concentration of the entire breeding pond; then use the sensitive spectral index of the chlorophyll a concentration of the entire breeding pond as the model input feature, import it into the chlorophyll a concentration estimation model trained in step d, and draw a spatial distribution map of the chlorophyll a concentration of the entire breeding pond; f. Based on step c, step d and step e, calculate the sensitive spectral index of the dissolved oxygen concentration of the entire breeding pond, use the sensitive spectral index of the dissolved oxygen concentration of the entire breeding pond and the spatial distribution map of chlorophyll a concentration as model input features, import them into the dissolved oxygen concentration estimation model trained in step d, and draw the spatial distribution map of the dissolved oxygen concentration of the entire breeding pond.

2. The method for monitoring dissolved oxygen concentration in freshwater aquaculture ponds based on multispectral imaging according to claim 1, characterized in that: In step a, the water sample data of the local breeding pond and the multispectral image data collected by the drone should be collected between 9:00 and 10:00 a.m. or between 15:00 and 16:00 p.m.; When collecting water sample data in local aquaculture ponds, the sampling depth is 30 cm underwater, and the water sample data collection method is the five-point mixing method. First, use an opaque container to mix the collected water samples evenly; then put the water samples into 2L and 500ml opaque water bottles respectively, of which 2L water samples are used to detect chlorophyll a concentration, and 500ml water samples are used to detect dissolved oxygen concentration. 2ml of magnesium carbonate suspension is dripped into the 2L sampling water bottle to prevent acidification from causing pigment dissolution; after the collection is completed, the 2L and 500m water bottles are promptly placed in an insulated box with ice bags for low temperature and light protection, and transported, and sent to the laboratory for inspection within 8 hours; the number of water sample data collected should be no less than 60.

3. The method for monitoring dissolved oxygen concentration in freshwater aquaculture ponds based on multispectral imaging according to claim 2, characterized in that: In the modeling process in step d, if the chlorophyll a concentration estimation model and the dissolved oxygen concentration estimation model cannot meet the evaluation criteria, it is necessary to establish more spectral indices and use more correlated sensitive spectral indices as model input features to improve the performance of the estimation model, and check whether there are outliers in the sample data and eliminate the outliers until the evaluation criteria are met.

Citation Information

Cited By

  • Method for monitoring water quality of aquaculture water area

    CN120558872A

  • Fish pond dissolved oxygen detection circuit

    CN120558923A