A method and related device for monitoring and analyzing aquatic areas based on machine vision

Abstract

The present invention discloses a method and related device for monitoring and analyzing aquatic areas based on machine vision, and relates to the field of image processing technology. The method comprises: preprocessing a target monitoring image of an aquatic area; performing color space decomposition on the monitoring image to obtain three channel images to generate a frequency domain image, and then generating a first target monitoring image feature; performing key point detection on the monitoring image, and generating a second target monitoring image feature based on the channel image combined with the target key point; fusing the first and second target monitoring image features, and then calculating the water quality value using a vector regression model; extracting color feature values ​​to obtain algae type information and algae distribution area information, and then generating an algae pollution analysis result; and performing a comprehensive situation analysis based on the water quality value and the algae pollution analysis result. The present invention improves the reliability of monitoring image data analysis in aquatic areas, so that its monitoring image data analysis can achieve a more ideal effect.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of image processing, and in particular to a method and related devices for monitoring and analyzing aquatic areas based on machine vision. Background Art

[0002] In the process of shrimp farming in aquatic areas, it is necessary to monitor the aquatic areas where the shrimps are farmed, and then analyze the relevant conditions of the aquatic areas based on the monitoring images, so as to facilitate timely adjustments to the aquatic areas. At present, in the data analysis of monitoring images of aquatic areas, the monitoring images are usually extracted by feature extractors to analyze the water quality of the aquatic areas. However, since the change in illumination will cause a large change in the water surface, the image features extracted by the feature extractor contain more redundant pixels, resulting in low accuracy of water quality analysis in aquatic areas. At the same time, due to the large breeding density and the large amount of shrimp feed fed, the water in the breeding area is more prone to eutrophication, making the water conditions in the aquatic areas more complex. However, the feature extractor fails to extract the target features well in the monitoring images of complex water conditions, resulting in unreliable subsequent water quality analysis results. At the same time, only a single feature is used to analyze the water quality value of the aquatic area, resulting in the accuracy of the final water quality value calculation failing to achieve the expected effect. Inaccurate water quality analysis will lead to the inability to adjust the water quality in a timely and accurate manner, which may cause health problems in farmed shrimp. In the data analysis of aquatic areas, the impact of algae in aquatic areas on farmed shrimp is rarely considered. On the one hand, the harmful algae in aquatic areas will consume a large amount of nutrients in the water and affect the growth of shrimp. On the other hand, these harmful algae will attach to the gills and cheeks of shrimp, seriously affecting their activities and feeding. Therefore, ignoring the data analysis of algae in aquatic areas will make the situation analysis of aquatic areas incomplete, resulting in the monitoring image data analysis of aquatic areas failing to achieve the desired effect. Summary of the invention

[0003] The purpose of the present invention is to overcome the shortcomings of the prior art. The present invention provides a method and related devices for monitoring and analyzing aquatic areas based on machine vision, which improves the reliability of monitoring image data analysis in aquatic areas and enables monitoring image data analysis in aquatic areas to achieve more ideal results.

[0004] In order to solve the above technical problems, the present invention provides a method for monitoring and analyzing aquatic areas based on machine vision, the method comprising:

[0005] Acquire a target monitoring image of the aquatic area, and preprocess the target monitoring image to obtain a preprocessed target monitoring image;

[0006] Performing color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image, and a V channel image corresponding to the preprocessed target monitoring image, generating a corresponding frequency domain image based on the H channel image, the S channel image, and the V channel image, performing feature extraction based on the frequency domain image, and obtaining a first target monitoring image feature;

[0007] Perform key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain corresponding target key points, and generate a second target monitoring image feature based on the H channel image, the S channel image, and the V channel image in combination with the target key points;

[0008] The first target monitoring image feature and the second target monitoring image feature are fused to obtain a target fusion feature, and a water quality value of the aquatic area is calculated using a vector regression model based on the target fusion feature;

[0009] Extracting color feature values ​​based on the preprocessed target monitoring image, analyzing algae types and algae distribution areas based on the color feature values ​​to obtain algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results;

[0010] Based on the water quality values ​​and algae pollution analysis results, a comprehensive situation analysis of the aquatic area is performed to obtain a comprehensive situation analysis result, and based on the comprehensive situation analysis result, it is determined whether it is necessary to issue an early warning information.

[0011] Optionally, preprocessing the target monitoring image to obtain the preprocessed target monitoring image includes:

[0012] Performing denoising processing on the target monitoring image to obtain a denoised target monitoring image;

[0013] The denoised target monitoring image is subjected to image enhancement processing to obtain a preprocessed target monitoring image.

[0014] Optionally, performing color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image, and a V channel image corresponding to the preprocessed target monitoring image, generating a corresponding frequency domain image based on the H channel image, the S channel image, and the V channel image, and performing feature extraction based on the frequency domain image to obtain a first target monitoring image feature, includes:

[0015] The preprocessed target monitoring image is converted from the RGB color space to the HSV color space based on the rgb2hsv function, and the preprocessed target monitoring image is subjected to color channel decomposition processing based on the HSV color space to obtain an H channel image, an S channel image, and a V channel image corresponding to the preprocessed target monitoring image;

[0016] Performing Fourier transform on the H channel image, the S channel image and the V channel image to obtain a plurality of frequency domain images corresponding to the H channel image, the S channel image and the V channel image;

[0017] Extracting features of the frequency domain image based on a preset color moment to obtain image color component features;

[0018] Acquire the image center of each frequency domain image, divide the image center of each frequency domain image into a plurality of concentric circles using a preset radius, and divide the plurality of concentric circles into a plurality of sector-shaped areas based on a preset angle;

[0019] Obtain pixel values ​​of all pixel points in each sector area, calculate the variance of all pixel values ​​in each sector area based on the pixel values ​​of all pixel points, and generate a first target monitoring image feature based on the variance of all pixel values ​​in each sector area and the image color component feature.

[0020] Optionally, performing key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain corresponding target key points, and generating a second target monitoring image feature based on the H channel image, the S channel image, and the V channel image in combination with the target key points, includes:

[0021] Constructing a Gaussian pyramid based on a Gaussian function, and generating a differential Gaussian scale space based on the Gaussian pyramid using a differential Gaussian scale space function;

[0022] Based on the differential Gaussian scale space, an extreme point search process is performed on the preprocessed target monitoring image to obtain a corresponding number of extreme points, and the number of extreme points are screened based on a preset screening rule, and the screened number of extreme points are used as target key points corresponding to the preprocessed target monitoring image;

[0023] A neighborhood window is set based on the target key point, a color mean of all pixels of the H channel image, the S channel image, and the V channel image within the neighborhood window is calculated, and a color feature is generated based on the color mean of all pixels within the neighborhood window;

[0024] Based on the target key points, the H channel image, the S channel image and the V channel image are divided into regions to obtain a plurality of pixel regions, and a corresponding directional gradient histogram is extracted for each pixel region;

[0025] A second target monitoring image feature is generated based on the directional gradient histogram of all pixel regions and the color feature.

[0026] Optionally, fusing the first target monitoring image feature and the second target monitoring image feature to obtain a target fusion feature, and calculating the water quality value of the aquatic area using a vector regression model based on the target fusion feature, includes:

[0027] Generate a first target feature vector based on the first target monitoring image feature, generate a second target feature vector based on the second target monitoring image feature, and concatenate the first target feature vector and the second target feature vector to obtain a target fusion feature;

[0028] Inputting the target fusion feature into the vector regression model, and calculating the initial water quality value based on the vector regression model;

[0029] Calculating the optimization function value corresponding to the initial water quality value based on a preset water quality parameter optimization function, and optimizing the target fusion feature based on the optimization function value to obtain an optimized target fusion feature;

[0030] The optimization target fusion feature is input into the vector regression model, and the water quality value of the aquatic area is calculated based on the vector regression model.

[0031] Optionally, the performing of algae type and algae distribution area analysis based on the color feature value to obtain algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results, includes:

[0032] Extracting color contrast feature values ​​based on an algae image database, performing similarity analysis based on the color feature values ​​and the color contrast feature values ​​to obtain similarity analysis results, and performing algae type analysis based on the similarity analysis results to obtain algae type information;

[0033] Performing algae distribution area analysis based on the color feature value combined with the algae type information to obtain algae distribution area information;

[0034] Based on the algae type information and the algae distribution area information, data simulation analysis is performed using the algae growth trend data to obtain algae growth simulation analysis data;

[0035] Based on the algae type information and the algae distribution area information combined with the algae growth simulation analysis data, an algae pollution analysis is performed to obtain an algae pollution analysis result.

[0036] Optionally, the performing a comprehensive situation analysis on the aquatic area based on the water quality value and the algae pollution analysis result to obtain a comprehensive situation analysis result includes:

[0037] The water quality values ​​and algae pollution analysis results are compared with preset expected values ​​to obtain comparison results, and a comprehensive situation analysis of the aquatic area is performed based on the comparison results to obtain a comprehensive situation analysis result.

[0038] In addition, the present invention also provides a device for monitoring and analyzing aquatic areas based on machine vision, the device comprising:

[0039] Monitoring image acquisition module: used to acquire the target monitoring image of the aquatic area, and pre-process the target monitoring image to obtain the pre-processed target monitoring image;

[0040] A first image feature acquisition module is used to perform color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image, and generate a corresponding frequency domain image based on the H channel image, the S channel image and the V channel image, and perform feature extraction based on the frequency domain image to obtain a first target monitoring image feature;

[0041] The second image feature acquisition module is used to perform key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain the corresponding target key points, and generate the second target monitoring image features based on the H channel image, the S channel image and the V channel image in combination with the target key points;

[0042] Image feature fusion and water quality value calculation module: used to fuse the first target monitoring image feature and the second target monitoring image feature to obtain a target fusion feature, and calculate the water quality value of the aquatic area using a vector regression model based on the target fusion feature;

[0043] Algae pollution analysis module: used for extracting color feature values ​​based on the preprocessed target monitoring image, performing algae type and algae distribution area analysis based on the color feature values, obtaining algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results;

[0044] Comprehensive analysis module: used to conduct a comprehensive situation analysis on the aquatic area based on the water quality value and the algae pollution analysis results, obtain a comprehensive situation analysis result, and determine whether it is necessary to issue an early warning information based on the comprehensive situation analysis result.

[0045] In addition, the present invention also provides an electronic device, which includes a processor and a memory, wherein the memory is used to store instructions, and the processor is used to call the instructions in the memory so that the electronic device executes the above-mentioned machine vision-based aquatic area monitoring and analysis method.

[0046] In addition, the present invention also provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed on an electronic device, the electronic device executes the above-mentioned machine vision-based aquatic area monitoring and analysis method.

[0047] In an embodiment of the present invention, the H channel image, S channel image and V channel image of the HSV color space can more intuitively express the image color features, and at the same time, the several frequency domain images generated by the H channel image, S channel image and V channel image are divided into sector areas to calculate the pixel value variance, and then the first target monitoring image features are generated according to the variance of all pixel values ​​in each sector area and the image color component features, so that the accuracy of image feature extraction can be guaranteed even in more complex water conditions. Based on the H channel image, S channel image and V channel image combined with the target key points obtained by the differential Gaussian scale space, the second target monitoring image features are generated, which can effectively eliminate redundant pixels in the image features and improve the reliability of target feature extraction. Based on the target fusion features generated by the first target monitoring image features and the second target monitoring image features, the water quality value of the aquatic area is calculated using a vector regression model. By combining different features together, the accuracy and robustness of the model can be improved, and a more accurate water quality value can be obtained. At the same time, the reliability of water quality value calculation is further improved by combining the vector regression model. The algae pollution analysis results obtained from the algae type information and algae distribution area information are also considered in the data analysis of the aquatic area, making the data analysis of the aquatic area more comprehensive and more targeted. The reliability of the monitoring image data analysis of the aquatic area is improved, so that the monitoring image data analysis of the aquatic area can achieve more ideal results, thereby ensuring the timely and accurate treatment of water quality and the healthy growth of shrimp. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 creative work.

[0049] Figure 1 is a flow chart of a method for monitoring and analyzing aquatic areas based on machine vision in an embodiment of the present invention;

[0050] Figure 2 Schematic diagram of the structure of the aquatic area monitoring and analysis device based on machine vision in an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments 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 creative work are within the scope of protection of the present invention.

[0053] Embodiment 1

[0054] See also Figure 1 , Figure 1 It is a flow chart of the aquatic area monitoring and analysis method based on machine vision in an embodiment of the present invention.

[0055] like Figure 1 As shown, a method for monitoring and analyzing aquatic areas based on machine vision, the method comprising:

[0056] S11: Acquire a target monitoring image of the aquatic area, and preprocess the target monitoring image to obtain a preprocessed target monitoring image;

[0057] In the specific implementation process of the present invention, the preprocessing of the target monitoring image to obtain the preprocessed target monitoring image includes: denoising the target monitoring image to obtain the denoised target monitoring image; performing image enhancement processing on the denoised target monitoring image to obtain the preprocessed target monitoring image.

[0058] Specifically, the target monitoring image of the aquatic area is collected by the monitoring equipment at a preset time interval. The aquatic area is the aquaculture area where the shrimp is located. The monitoring image of the aquatic area where the shrimp is located is collected so that its specific situation can be analyzed in time. The target monitoring image is denoised, and the target monitoring image is preliminarily denoised using a two-dimensional discrete cosine transform domain to obtain the target monitoring image after the preliminary denoising, and the preset energy parameters of the target monitoring image after the preliminary denoising are obtained. According to the preset energy parameters, the residual amount of the central pixel of each unit area of ​​the target monitoring image after the preliminary denoising is obtained, and the weight matrix of each unit area is calculated based on the residual amount. The target monitoring image after the preliminary denoising is further denoised based on the weight matrix to obtain the target monitoring image after the preliminary denoising. The denoised target monitoring image is enhanced, the object field of the denoised target monitoring image is modulated in the frequency domain, the frequency domain is Fourier transformed, the imaginary part of the frequency domain after the Fourier transform is taken, the absolute value of the imaginary part is calculated, the power value is calculated based on the absolute value of the imaginary part after summing along the frequency, the image is enhanced based on the power value, and the image preprocessing is completed after the image enhancement is completed, and the preprocessed target monitoring image is obtained.

[0059] S12: performing color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image, and a V channel image corresponding to the preprocessed target monitoring image, and generating a corresponding frequency domain image based on the H channel image, the S channel image, and the V channel image, performing feature extraction based on the frequency domain image to obtain a first target monitoring image feature;

[0060] In the specific implementation process of the present invention, the preprocessed target monitoring image is subjected to color space decomposition processing to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image, and a corresponding frequency domain image is generated based on the H channel image, the S channel image and the V channel image, and a feature is extracted based on the frequency domain image to obtain a first target monitoring image feature, including: converting the preprocessed target monitoring image from an RGB color space to an HSV color space based on an rgb2hsv function, and performing color channel decomposition processing on the preprocessed target monitoring image based on the HSV color space to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image; The H channel image, the S channel image and the V channel image are subjected to Fourier transformation to obtain a plurality of frequency domain images corresponding to the H channel image, the S channel image and the V channel image; feature extraction is performed on the frequency domain image based on a preset color moment to obtain image color component features; the image center of each frequency domain image is obtained, a plurality of concentric circles are divided into a plurality of concentric circles based on a preset radius based on the image center of each frequency domain image, and a plurality of sector-shaped areas are divided in the plurality of concentric circles based on a preset angle; pixel values ​​of all pixel points in each sector-shaped area are obtained, the variance of all pixel values ​​of each sector-shaped area is calculated based on the pixel values ​​of all pixel points, and the first target monitoring image feature is generated based on the variance of all pixel values ​​of each sector-shaped area and the image color component feature.

[0061] Specifically, the preprocessed target monitoring image is converted from the red-green-blue (RGB) color space to the hue-saturation-value (HSV) color space based on the rgb2hsv function. The rgb2hsv function is a function for realizing the conversion between the RGB color space and the HSB color space. The preprocessed target monitoring image is subjected to color channel decomposition processing based on the HSV color space. In the HSV color space, there are hue (Hue, H) channel, saturation (Saturation, S) channel and brightness (Value, V) channel respectively. The color channel decomposition is performed in the HSV color space to obtain images of each color channel, that is, to obtain the H channel image, S channel image and V channel image corresponding to the preprocessed target monitoring image. The image is converted from the RGB space to the HSV color space, which can very intuitively express the brightness, hue and vividness of the image color, and facilitate the comparison between features. The H channel image, the S channel image and the V channel image are subjected to Fourier transform. The Fourier transform may adopt a two-dimensional discrete Fourier transform to respectively obtain the sizes, the horizontal and vertical coordinates of the pixels and the frequency variables of the H channel image, the S channel image and the V channel image. The Fourier transform may be performed according to the sizes, the horizontal and vertical coordinates of the pixels and the frequency variables of the H channel image, the S channel image and the V channel image using a two-dimensional discrete Fourier transform formula to obtain a plurality of frequency domain images corresponding to the H channel image, the S channel image and the V channel image, that is, each channel image corresponds to a frequency domain image. Based on the preset color moment, the frequency domain image is feature extracted to obtain the number of pixels and the color component of each pixel of the frequency domain image. The color moment is calculated using the preset first to third order color moment calculation formula according to the number of pixels and the color component of each pixel of the frequency domain image to obtain several color moments, wherein the first order color moment uses the first order origin matrix to represent the overall brightness of the image, the second order color moment uses the square root of the second order center moment to represent the color distribution of the image, and the third order color moment uses the cube root of the third order center moment to represent the symmetry of the color distribution of the image. The several color moments are used as the image color component features to obtain the image color component features. The image center of each frequency domain image is obtained, the frequency domain rate image is converted from the rectangular coordinate system to the polar coordinate system, and the image center of the frequency domain rate image is obtained in the polar coordinate system; based on the image center of each frequency domain image, several concentric circles are divided using the preset radius, and several sector areas are divided in the several concentric circles based on the preset angle, and the sector areas are equiangular sectors. The pixel values ​​of all pixels in each sector area are obtained, the average pixel value is calculated according to the pixel values ​​of all pixels in each sector area, and the variance of all pixel values ​​in each sector area is calculated based on the pixel values ​​of all pixels using the average pixel value combined with the variance calculation formula.The first target monitoring image feature is generated based on the variance of all pixel values ​​in each sector area and the image color component feature. The variance of all pixel values ​​in each sector area is used as the pixel statistical feature of the image. Combined with several color moments in the image color component feature, the first target monitoring image feature can be made more accurate, and the accuracy of image feature extraction can be guaranteed even in more complex water conditions.

[0062] S13: performing key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain corresponding target key points, and generating a second target monitoring image feature based on the H channel image, the S channel image, and the V channel image in combination with the target key points;

[0063] In the specific implementation process of the present invention, the target monitoring image after preprocessing is detected based on the differential Gaussian scale space to obtain corresponding target key points, and the second target monitoring image features are generated based on the H channel image, the S channel image and the V channel image in combination with the target key points, including: constructing a Gaussian pyramid based on a Gaussian function, and generating a differential Gaussian scale space based on the Gaussian pyramid using a differential Gaussian scale space function; performing extreme point search processing on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain a number of corresponding extreme points, and screening a number of extreme points based on a preset screening rule, and screening the screened extreme points. The method comprises the following steps: selecting several extreme value points after the preprocessing as target key points corresponding to the preprocessed target monitoring image; setting a neighborhood window based on the target key points, calculating the color mean of all pixel points of the H channel image, the S channel image and the V channel image in the neighborhood window, and generating color features based on the color mean of all pixel points in the neighborhood window; performing region division on the H channel image, the S channel image and the V channel image based on the target key points to obtain several pixel regions, and extracting a corresponding directional gradient histogram for each pixel region; and generating a second target monitoring image feature based on the directional gradient histograms of all pixel regions and the color features.

[0064] Specifically, a Gaussian pyramid is constructed based on a Gaussian function, and a Gaussian function with different standard deviations is used to perform convolution processing on the preprocessed target monitoring image. During the convolution processing, the Gaussian scale space is divided into several groups by a Gaussian scale space function, and row convolution and column convolution are performed in each group respectively. A row convolution module based on a programmable gate array is used to perform row convolution processing to obtain row convolution image information, and the row convolution image information is converted according to a row-column conversion module to obtain the converted row convolution image information, and then a column convolution processing is performed based on a column convolution module of a programmable gate array to obtain column convolution image information, and a Gaussian pyramid is formed according to the row convolution image information and the column convolution image information. Based on the Gaussian pyramid, a differential Gaussian scale space is generated using a differential Gaussian scale space function, and the difference of adjacent scale layers is obtained in the Gaussian scale space based on the Gaussian pyramid, and a differential Gaussian scale space is generated according to the difference of adjacent scale layers combined with the differential Gaussian scale space function. Based on the differential Gaussian scale space, an extreme point search process is performed on the preprocessed target monitoring image. The pixel point in the preprocessed target monitoring image is compared with the adjacent points of the pixel point in the image domain of the same scale through the differential Gaussian scale space, and the pixel point is compared with all adjacent points in the scale domain of the adjacent scale space in the same scale space. When the pixel point is larger or smaller than all adjacent points, it is an extreme point, and then a corresponding number of extreme points are obtained. Based on the preset screening rules, several extreme points are screened, that is, points with low contrast and unstable edge response points are removed. The objective function value is calculated for each extreme point based on the objective function. If the objective function value does not reach the preset interval, it is judged that the contrast of the extreme point is low, and the extreme point is removed. The Gaussian difference operator is used to calculate the principal curvature of each extreme point in the direction across the edge and the principal curvature in the direction perpendicular to the edge. The principal curvature in the direction across the edge and the principal curvature in the direction perpendicular to the edge are compared. If the principal curvature in the direction across the edge is greater than the principal curvature in the direction perpendicular to the edge, the extreme point is judged to be an unstable edge corresponding point, and the extreme point is removed until all low-contrast points and unstable edge response points are removed. The retained extreme points are obtained, which are the screened extreme points. The screened extreme points are used as target key points corresponding to the preprocessed target monitoring image. A neighborhood window is set based on the target key point, that is, a neighborhood window is set with the target key point as the center, the color mean of all pixels of the H channel image, the S channel image and the V channel image in the neighborhood window is calculated, the pixel values ​​of all pixels of the H channel image, the S channel image and the V channel image in the neighborhood window are obtained, and the pixel average value is calculated by the pixel values ​​of all pixels in the neighborhood window, that is, the color mean. A color feature is generated based on the color mean of all pixels in the neighborhood window, and the color mean of all pixels in the neighborhood window is used as its color statistical feature.Based on the target key points, the H channel image, S channel image and V channel image are divided into regions, the pixel region division size is set according to the target key points, and the H channel image, S channel image and V channel image are divided into regions according to the pixel region division size to obtain a plurality of pixel regions. The corresponding directional gradient histogram is extracted for each pixel region, the gradient direction and gradient value of each pixel in each pixel region are calculated, and the feature description of the directional gradient histogram is formed according to the gradient direction and gradient value of each pixel. The directional gradient histogram based on all pixel regions and the color feature constitutes the second target monitoring image feature. Based on the H channel image, S channel image and V channel image combined with the target key points obtained by the differential Gaussian scale space, the second target monitoring image feature is generated, which can effectively eliminate redundant pixels in the image features and improve the reliability of target feature extraction.

[0065] S14: fusing the first target monitoring image feature and the second target monitoring image feature to obtain a target fusion feature, and calculating the water quality value of the aquatic area using a vector regression model based on the target fusion feature;

[0066] In the specific implementation process of the present invention, the first target monitoring image feature and the second target monitoring image feature are fused to obtain a target fusion feature, and the water quality value of the aquatic area is calculated using a vector regression model based on the target fusion feature, including: generating a first target feature vector based on the first target monitoring image feature, generating a second target feature vector based on the second target monitoring image feature, and splicing the first target feature vector and the second target feature vector to obtain a target fusion feature; inputting the target fusion feature into the vector regression model, and calculating the initial water quality value based on the vector regression model; calculating the optimization function value corresponding to the initial water quality value based on a preset water quality parameter optimization function, and optimizing the target fusion feature based on the optimization function value to obtain an optimized target fusion feature; inputting the optimized target fusion feature into the vector regression model, and calculating the water quality value of the aquatic area based on the vector regression model.

[0067] Specifically, a first target feature vector is generated based on the first target monitoring image feature, and the first target monitoring image feature is numerically processed based on a standardized method. The purpose of the numerical processing is to convert the feature into a numerical form, and then a one-dimensional feature vector is generated based on the first target monitoring image feature after the numerical processing, which is the first target feature vector. The purpose of generating the feature vector is to integrate the image feature information, reduce the dimension of the feature, and reduce the calculation complexity of the subsequent model; a second target feature vector is generated based on the second target monitoring image feature, and the second target monitoring image feature is numerically processed based on a standardized method. The purpose of the numerical processing is to convert the feature into a numerical form, and then a two-dimensional feature vector is generated based on the second target monitoring image feature after the numerical processing, which is the second target feature vector. The first target feature vector and the second target feature vector are spliced ​​to obtain a target fusion feature. The feature vectors are fused by splicing, which expands the vector space and better retains the semantic information of each latent vector dimension. By combining different features together, more comprehensive feature information can be obtained, which can improve the calculation accuracy of the model. At the same time, because different features can capture different information, feature fusion can make the model more robust. The target fusion feature is input into the vector regression model, and the initial water quality value is calculated based on the vector regression model. The vector regression model is:

[0068]

[0069] Where W is the water quality value, A is the bias term, δ is the water quality optimization factor, N is the number of features, i is the feature number in the target fusion feature, exp is the exponential function, S is the feature mean of the target fusion feature, S i is the vector value of the i-th feature in the target fusion feature, σ is the correction coefficient; the bias term is a parameter in the linear model, and the bias term represents the amount of translation on the linear combination of the input features. The water quality optimization factor can be obtained from the historical water quality data in the database. The optimization function value corresponding to the initial water quality value is calculated based on the preset water quality parameter optimization function. The preset water quality parameter optimization function can select the mean square error function and the absolute error function, etc., and the target fusion feature is optimized based on the optimization function value, that is, the optimization function value is added to each feature in the target fusion feature to obtain the optimized target fusion feature. The optimized target fusion feature is input into the vector regression model, and the water quality value of the aquatic area is calculated based on the vector regression model. After the optimization function value is added to the target fusion feature, the error effect is further reduced. The accuracy and reliability of the water quality value of the final aquatic area calculated using the vector regression model according to the optimized target fusion feature will be higher.

[0070] S15: extracting color feature values ​​based on the preprocessed target monitoring image, analyzing algae types and algae distribution areas based on the color feature values ​​to obtain algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results;

[0071] In the specific implementation process of the present invention, the algae type and algae distribution area analysis is performed based on the color feature value to obtain algae type information and algae distribution area information, and algae pollution analysis is performed based on the algae type information and algae distribution area information to obtain algae pollution analysis results, including: extracting color contrast feature values ​​based on an algae image database, performing similarity analysis based on the color feature values ​​and color contrast feature values ​​to obtain similarity analysis results, and performing algae type analysis based on the similarity analysis results to obtain algae type information; performing algae distribution area analysis based on the color feature values ​​combined with the algae type information to obtain algae distribution area information; performing data simulation analysis using algae growth trend data based on the algae type information and algae distribution area information to obtain algae growth simulation analysis data; performing algae pollution analysis based on the algae type information and algae distribution area information combined with the algae growth simulation analysis data to obtain algae pollution analysis results.

[0072] Specifically, the color average of the preprocessed target monitoring image is calculated, the color channel value of each pixel in the preprocessed target monitoring image is obtained, and the color characteristic value is obtained based on the color average and the color channel value of each pixel using the color channel correction coefficient combined with the color correction coefficient. The color contrast characteristic value is extracted based on the algae image database, and the similarity analysis is performed based on the color characteristic value and the color contrast characteristic value. Since the colors formed by different algae on the water surface are different, the algae type information can be quickly obtained by performing similarity analysis through the color characteristic value and the color contrast characteristic value. The similarity analysis can be performed by using the similarity comparison software to obtain the similarity analysis result. The algae type analysis is performed based on the similarity analysis result, and the algae type corresponding to the color contrast characteristic value with the maximum similarity can be obtained through the similarity analysis result, from which the algae type information can be obtained. The algae distribution area analysis is performed based on the color characteristic value combined with the algae type information, and the color distribution area is calculated in the preprocessed target monitoring image based on the color characteristic value. The colors formed by different algae types are also different, so the algae distribution area information can be obtained according to the color distribution area combined with the algae type information. Based on the algae type information and algae distribution area information, data simulation analysis is performed using algae growth trend data. The algae type information can be used to determine the type of algae, the algae distribution area information can be used to determine the distribution quantity of the corresponding algae, and the algae growth trend data can be used to determine the reproduction rate of the corresponding algae. The algae growth trend information can be obtained in the algae database. The algae type information, algae distribution area information, and algae growth trend data are imported into the algae growth simulation model to obtain algae growth simulation analysis data. Based on the algae type information and algae distribution area information combined with the algae growth simulation analysis data, algae pollution analysis is performed. The reproduction rate and reproduction quantity of the corresponding type of algae in the aquaculture area per unit time can be obtained through the algae type information and the algae distribution area information combined with the algae growth simulation analysis data, thereby obtaining the algae pollution analysis results in the shrimp aquaculture area.

[0073] S16: Conduct a comprehensive situation analysis on the aquatic area based on the water quality value and the algae pollution analysis result to obtain a comprehensive situation analysis result, and determine whether it is necessary to issue an early warning message based on the comprehensive situation analysis result.

[0074] In the specific implementation process of the present invention, a comprehensive situation analysis is performed on the aquatic area based on the water quality values ​​and the algae pollution analysis results to obtain a comprehensive situation analysis result, including: comparing the water quality values ​​and the algae pollution analysis results with preset expected values ​​to obtain a comparison result, and performing a comprehensive situation analysis on the aquatic area based on the comparison result to obtain a comprehensive situation analysis result.

[0075] Specifically, the water quality value and algae pollution analysis result are compared with the preset expected value, which is the water quality value and algae distribution value expected in the shrimp aquaculture area, so that the difference between the current aquaculture area and the preset expected value can be known, that is, the comparison result is obtained. Based on the comparison result, a comprehensive situation analysis is performed on the aquaculture area. According to the comparison result, the algae pollution level and water quality in the aquaculture area can be analyzed to see whether they will affect the healthy growth of shrimp, and a comprehensive situation analysis result can be obtained. Based on the comprehensive situation analysis result, it is determined whether it is necessary to issue an early warning message. If it is determined that the algae pollution level and water quality in the aquaculture area will affect the healthy growth of shrimp, an early warning message is issued to notify relevant personnel to handle the problem, such as changing water and removing algae.

[0076] In an embodiment of the present invention, the H channel image, S channel image and V channel image of the HSV color space can more intuitively express the image color features, and at the same time, the several frequency domain images generated by the H channel image, S channel image and V channel image are divided into sector areas to calculate the pixel value variance, and then the first target monitoring image features are generated according to the variance of all pixel values ​​in each sector area and the image color component features, so that the accuracy of image feature extraction can be guaranteed even in more complex water conditions. Based on the H channel image, S channel image and V channel image combined with the target key points obtained by the differential Gaussian scale space, the second target monitoring image features are generated, which can effectively eliminate redundant pixels in the image features and improve the reliability of target feature extraction. Based on the target fusion features generated by the first target monitoring image features and the second target monitoring image features, the water quality value of the aquatic area is calculated using a vector regression model. By combining different features together, the accuracy and robustness of the model can be improved, and a more accurate water quality value can be obtained. At the same time, the reliability of water quality value calculation is further improved by combining the vector regression model. The algae pollution analysis results obtained from the algae type information and algae distribution area information are also considered in the data analysis of the aquatic area, making the data analysis of the aquatic area more comprehensive and more targeted. The reliability of the monitoring image data analysis of the aquatic area is improved, so that the monitoring image data analysis of the aquatic area can achieve more ideal results, thereby ensuring the timely and accurate treatment of water quality and the healthy growth of shrimp.

[0077] Embodiment 2

[0078] See also Figure 2 , Figure 2 It is a schematic diagram of the structural composition of a device for monitoring and analyzing aquatic areas based on machine vision in an embodiment of the present invention.

[0079] like Figure 2 As shown, a device for monitoring and analyzing aquatic areas based on machine vision, the device comprises:

[0080] Monitoring image acquisition module 21: used to acquire a target monitoring image of the aquatic area, and pre-process the target monitoring image to obtain a pre-processed target monitoring image;

[0081] The first image feature acquisition module 22 is used to perform color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image, and generate a corresponding frequency domain image based on the H channel image, the S channel image and the V channel image, and perform feature extraction based on the frequency domain image to obtain a first target monitoring image feature;

[0082] The second image feature acquisition module 23 is used to perform key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain the corresponding target key points, and generate the second target monitoring image features based on the H channel image, the S channel image and the V channel image in combination with the target key points;

[0083] Image feature fusion and water quality value calculation module 24: used to fuse the first target monitoring image feature and the second target monitoring image feature to obtain a target fusion feature, and calculate the water quality value of the aquatic area using a vector regression model based on the target fusion feature;

[0084] Algae pollution analysis module 25: used for extracting color feature values ​​based on the preprocessed target monitoring image, performing algae type and algae distribution area analysis based on the color feature values ​​to obtain algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results;

[0085] Comprehensive analysis module 26: used to conduct a comprehensive situation analysis on the aquatic area based on the water quality value and the algae pollution analysis result, obtain a comprehensive situation analysis result, and determine whether it is necessary to issue an early warning information based on the comprehensive situation analysis result.

[0086] In the specific implementation process of the present invention, the specific implementation method of the device item can refer to the implementation method of the above-mentioned method item, which will not be repeated here.

[0087] In an embodiment of the present invention, the H channel image, S channel image and V channel image of the HSV color space can more intuitively express the image color features, and at the same time, the several frequency domain images generated by the H channel image, S channel image and V channel image are divided into sector areas to calculate the pixel value variance, and then the first target monitoring image features are generated according to the variance of all pixel values ​​in each sector area and the image color component features, so that the accuracy of image feature extraction can be guaranteed even in more complex water conditions. Based on the H channel image, S channel image and V channel image combined with the target key points obtained by the differential Gaussian scale space, the second target monitoring image features are generated, which can effectively eliminate redundant pixels in the image features and improve the reliability of target feature extraction. Based on the target fusion features generated by the first target monitoring image features and the second target monitoring image features, the water quality value of the aquatic area is calculated using a vector regression model. By combining different features together, the accuracy and robustness of the model can be improved, and a more accurate water quality value can be obtained. At the same time, the reliability of water quality value calculation is further improved by combining the vector regression model. The algae pollution analysis results obtained from the algae type information and algae distribution area information are also considered in the data analysis of the aquatic area, making the data analysis of the aquatic area more comprehensive and more targeted. The reliability of the monitoring image data analysis of the aquatic area is improved, so that the monitoring image data analysis of the aquatic area can achieve more ideal results, thereby ensuring the timely and accurate treatment of water quality and the healthy growth of shrimp.

[0088] A computer-readable storage medium provided in an embodiment of the present invention stores a computer program on the computer-readable storage medium, and when the program is executed by a processor, a method for monitoring and analyzing aquatic areas based on machine vision in any of the above embodiments is implemented. The computer-readable storage medium includes, but is not limited to, any type of disk (including floppy disk, hard disk, optical disk, CD-ROM, and magneto-optical disk), ROM (Read-Only Memory), RAM (Random Access Memory), EPROM (Erasable Programmable Read-Only Memory), EEPROM (Electrically Erasable Programmable Read-Only Memory), flash memory, magnetic card or optical card. That is, the storage device includes any medium that can store or transmit information in a readable form by a device (e.g., a computer, a mobile phone), which can be a read-only memory, a disk or an optical disk, etc.

[0089] Embodiment 3

[0090] See also Figure 3 , Figure 3 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention.

[0091] The embodiment of the present invention further provides an electronic device, such as Figure 3 As shown, the electronic device includes a memory 31, a processor 33, and a computer program 32 stored in the memory 31 and executable on the processor 33. Those skilled in the art will appreciate that Figure 3 The electronic device shown does not constitute a limitation on all devices, and may include more or fewer components than shown, or combine certain components. The memory 31 can be used to store the computer program 32 and various functional modules, and the processor 33 runs the computer program 32 stored in the memory 31, thereby executing various functional applications and data processing of the device. The memory can be an internal memory or an external memory, or include both internal and external memories. The internal memory can include a read-only memory (ROM), a programmable ROM (PROM), an electrically programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a flash memory, or a random access memory. The external memory can include a hard disk, a floppy disk, a ZIP disk, a U disk, a magnetic tape, etc. The processor 33 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general processor may be a microprocessor, a single chip microcomputer or the processor 33 may also be any conventional processor, etc. The processor and memory disclosed in the present invention include but are not limited to these types of processors and memories. The processor and memory disclosed in the present invention are only examples and not limitations.

[0092] As an embodiment, the electronic device includes: one or more processors 33, a memory 31, and one or more computer programs 32, wherein the one or more computer programs 32 are stored in the memory 31 and are configured to be executed by the one or more processors 33, and the one or more computer programs 32 are configured to execute the machine vision-based aquatic area monitoring and analysis method in any of the above-mentioned embodiments. Please refer to the above-mentioned embodiments for the specific implementation process, which will not be repeated here.

[0093] In an embodiment of the present invention, the H channel image, S channel image and V channel image of the HSV color space can more intuitively express the image color features, and at the same time, the several frequency domain images generated by the H channel image, S channel image and V channel image are divided into sector areas to calculate the pixel value variance, and then the first target monitoring image features are generated according to the variance of all pixel values ​​in each sector area and the image color component features, so that the accuracy of image feature extraction can be guaranteed even in more complex water conditions. Based on the H channel image, S channel image and V channel image combined with the target key points obtained by the differential Gaussian scale space, the second target monitoring image features are generated, which can effectively eliminate redundant pixels in the image features and improve the reliability of target feature extraction. Based on the target fusion features generated by the first target monitoring image features and the second target monitoring image features, the water quality value of the aquatic area is calculated using a vector regression model. By combining different features together, the accuracy and robustness of the model can be improved, and a more accurate water quality value can be obtained. At the same time, the reliability of water quality value calculation is further improved by combining the vector regression model. The algae pollution analysis results obtained from the algae type information and algae distribution area information are also considered in the data analysis of the aquatic area, making the data analysis of the aquatic area more comprehensive and more targeted. The reliability of the monitoring image data analysis of the aquatic area is improved, so that the monitoring image data analysis of the aquatic area can achieve more ideal results, thereby ensuring the timely and accurate treatment of water quality and the healthy growth of shrimp.

[0094] In addition, the above is a detailed introduction to a machine vision-based aquatic area monitoring and analysis method and related devices provided in an embodiment of the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A method for monitoring and analyzing aquatic areas based on machine vision, characterized in that: The method comprises: Acquire a target monitoring image of the aquatic area, and preprocess the target monitoring image to obtain a preprocessed target monitoring image; Performing color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image, and a V channel image corresponding to the preprocessed target monitoring image, generating a corresponding frequency domain image based on the H channel image, the S channel image, and the V channel image, performing feature extraction based on the frequency domain image, and obtaining a first target monitoring image feature; Perform key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain corresponding target key points, and generate a second target monitoring image feature based on the H channel image, the S channel image, and the V channel image in combination with the target key points; The first target monitoring image feature and the second target monitoring image feature are fused to obtain a target fusion feature, and a water quality value of the aquatic area is calculated using a vector regression model based on the target fusion feature; Extracting color feature values ​​based on the preprocessed target monitoring image, analyzing algae types and algae distribution areas based on the color feature values ​​to obtain algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results; Conducting a comprehensive situation analysis on the aquatic area based on the water quality value and the algae pollution analysis result to obtain a comprehensive situation analysis result, and judging whether it is necessary to issue an early warning information based on the comprehensive situation analysis result; Among them, the color space decomposition processing is performed on the preprocessed target monitoring image to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image, and a corresponding frequency domain image is generated based on the H channel image, the S channel image and the V channel image, and feature extraction is performed based on the frequency domain image to obtain a first target monitoring image feature, including: converting the preprocessed target monitoring image from the RGB color space to the HSV color space based on the rgb2hsv function, and performing color channel decomposition processing on the preprocessed target monitoring image based on the HSV color space to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image; performing Fourier transform on the H channel image, the S channel image and the V channel image to obtain a number of frequency domain images corresponding to the H channel image, the S channel image and the V channel image; performing feature extraction on the frequency domain image based on a preset color moment to obtain pixel numbers of the frequency domain image. The number and color component of each pixel of the frequency domain image are calculated, and the color moment is calculated by using a preset first to third order color moment calculation formula according to the number of pixels of the frequency domain image and the color component of each pixel, to obtain a number of color moments, wherein the first order color moment adopts a first order origin matrix, the second order color moment adopts the square root of the second order central moment, and the third order color moment adopts the cube root of the third order central moment, and the number of color moments are used as image color component features to obtain image color component features; the image center of each frequency domain image is obtained, and a number of concentric circles are divided into a number of concentric circles based on the image center of each frequency domain image using a preset radius, and a number of sector areas are divided in the number of concentric circles based on a preset angle; the pixel values ​​of all pixel points in each sector area are obtained, and the variance of all pixel values ​​in each sector area is calculated based on the pixel values ​​of all pixel points, and the variance of all pixel values ​​in each sector area is used as the pixel statistical feature of the image, and the first target monitoring image feature is generated based on the variance of all pixel values ​​in each sector area and the image color component feature; The method performs key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain corresponding target key points, and generates a second target monitoring image feature based on the H channel image, the S channel image and the V channel image in combination with the target key points, including: constructing a Gaussian pyramid based on a Gaussian function, generating a differential Gaussian scale space based on the Gaussian pyramid using a differential Gaussian scale space function, obtaining differences between adjacent scale layers in the Gaussian scale space based on the Gaussian pyramid, and generating a differential Gaussian scale space based on differences between adjacent scale layers in combination with the differential Gaussian scale space function; performing extreme point search processing on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain a number of corresponding extreme points, and screening a number of extreme points based on a preset screening rule, that is, removing points with low contrast and unstable edge response points, and calculating an objective function value for each extreme point based on the objective function. If the objective function value does not reach a preset interval, it is determined that the contrast of the extreme point is low, and the extreme point is removed, and the principal curvature and the relative curvature of each extreme point in the direction across the edge are calculated by a Gaussian difference operator. The principal curvature in the vertical edge direction is compared with the principal curvature in the cross-edge direction and the principal curvature in the vertical edge direction. If the principal curvature in the cross-edge direction is greater than the principal curvature in the vertical edge direction, the extreme point is judged to be an unstable edge response point, and the extreme point is removed until all points with low contrast and unstable edge response points are removed. The retained extreme points are obtained, that is, the screened extreme points, and the screened extreme points are used as target key points corresponding to the preprocessed target monitoring image; a neighborhood window is set based on the target key point, the color mean of all pixel points of the H channel image, the S channel image and the V channel image in the neighborhood window is calculated, and color features are generated based on the color mean of all pixel points in the neighborhood window; the H channel image, the S channel image and the V channel image are divided into regions based on the target key point to obtain several pixel regions, and the corresponding directional gradient histogram is extracted for each pixel region; the second target monitoring image feature is generated based on the directional gradient histogram of all pixel regions and the color feature.

2. The aquatic area monitoring and analysis method based on machine vision according to claim 1 is characterized in that: The preprocessing of the target monitoring image to obtain the preprocessed target monitoring image includes: Performing denoising processing on the target monitoring image to obtain a denoised target monitoring image; The denoised target monitoring image is subjected to image enhancement processing to obtain a preprocessed target monitoring image.

3. The aquatic area monitoring and analysis method based on machine vision according to claim 1 is characterized in that: The step of fusing the first target monitoring image feature and the second target monitoring image feature to obtain a target fusion feature, and calculating the water quality value of the aquatic area using a vector regression model based on the target fusion feature includes: Generate a first target feature vector based on the first target monitoring image feature, generate a second target feature vector based on the second target monitoring image feature, and concatenate the first target feature vector and the second target feature vector to obtain a target fusion feature; Inputting the target fusion feature into the vector regression model, and calculating the initial water quality value based on the vector regression model; Calculating the optimization function value corresponding to the initial water quality value based on a preset water quality parameter optimization function, and optimizing the target fusion feature based on the optimization function value to obtain an optimized target fusion feature; The optimization target fusion feature is input into the vector regression model, and the water quality value of the aquatic area is calculated based on the vector regression model.

4. The aquatic area monitoring and analysis method based on machine vision according to claim 1 is characterized in that: The algae type and algae distribution area analysis is performed based on the color feature value to obtain algae type information and algae distribution area information, and the algae pollution analysis is performed based on the algae type information and the algae distribution area information to obtain the algae pollution analysis result, including: Extracting color contrast feature values ​​based on an algae image database, performing similarity analysis based on the color feature values ​​and the color contrast feature values ​​to obtain similarity analysis results, and performing algae type analysis based on the similarity analysis results to obtain algae type information; Performing algae distribution area analysis based on the color feature value combined with the algae type information to obtain algae distribution area information; Based on the algae type information and the algae distribution area information, data simulation analysis is performed using the algae growth trend data to obtain algae growth simulation analysis data; Based on the algae type information and the algae distribution area information combined with the algae growth simulation analysis data, an algae pollution analysis is performed to obtain an algae pollution analysis result.

5. The aquatic area monitoring and analysis method based on machine vision according to claim 1 is characterized in that: The comprehensive situation analysis of the aquatic area is performed based on the water quality value and the algae pollution analysis result to obtain the comprehensive situation analysis result, including: The water quality values ​​and algae pollution analysis results are compared with preset expected values ​​to obtain comparison results, and a comprehensive situation analysis of the aquatic area is performed based on the comparison results to obtain a comprehensive situation analysis result.

6. A machine vision-based aquatic area monitoring and analysis device, characterized in that: The device comprises: Monitoring image acquisition module: used to acquire the target monitoring image of the aquatic area, and pre-process the target monitoring image to obtain the pre-processed target monitoring image; A first image feature acquisition module is used to perform color space decomposition processing on the preprocessed target monitoring image to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image, and generate a corresponding frequency domain image based on the H channel image, the S channel image and the V channel image, and perform feature extraction based on the frequency domain image to obtain a first target monitoring image feature; The second image feature acquisition module is used to perform key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain the corresponding target key points, and generate the second target monitoring image features based on the H channel image, the S channel image and the V channel image in combination with the target key points; Image feature fusion and water quality value calculation module: used to fuse the first target monitoring image feature and the second target monitoring image feature to obtain a target fusion feature, and calculate the water quality value of the aquatic area using a vector regression model based on the target fusion feature; Algae pollution analysis module: used for extracting color feature values ​​based on the preprocessed target monitoring image, performing algae type and algae distribution area analysis based on the color feature values, obtaining algae type information and algae distribution area information, and performing algae pollution analysis based on the algae type information and algae distribution area information to obtain algae pollution analysis results; Comprehensive analysis module: used to conduct a comprehensive situation analysis on the aquatic area based on the water quality value and the algae pollution analysis result, obtain a comprehensive situation analysis result, and determine whether it is necessary to issue an early warning information based on the comprehensive situation analysis result; Among them, the color space decomposition processing is performed on the preprocessed target monitoring image to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image, and a corresponding frequency domain image is generated based on the H channel image, the S channel image and the V channel image, and feature extraction is performed based on the frequency domain image to obtain a first target monitoring image feature, including: converting the preprocessed target monitoring image from the RGB color space to the HSV color space based on the rgb2hsv function, and performing color channel decomposition processing on the preprocessed target monitoring image based on the HSV color space to obtain an H channel image, an S channel image and a V channel image corresponding to the preprocessed target monitoring image; performing Fourier transform on the H channel image, the S channel image and the V channel image to obtain a number of frequency domain images corresponding to the H channel image, the S channel image and the V channel image; performing feature extraction on the frequency domain image based on a preset color moment to obtain pixel numbers of the frequency domain image. The number and color component of each pixel of the frequency domain image are calculated, and the color moment is calculated by using a preset first to third order color moment calculation formula according to the number of pixels of the frequency domain image and the color component of each pixel, to obtain a number of color moments, wherein the first order color moment adopts a first order origin matrix, the second order color moment adopts the square root of the second order central moment, and the third order color moment adopts the cube root of the third order central moment, and the number of color moments are used as image color component features to obtain image color component features; the image center of each frequency domain image is obtained, and a number of concentric circles are divided into a number of concentric circles based on the image center of each frequency domain image using a preset radius, and a number of sector areas are divided in the number of concentric circles based on a preset angle; the pixel values ​​of all pixel points in each sector area are obtained, and the variance of all pixel values ​​in each sector area is calculated based on the pixel values ​​of all pixel points, and the variance of all pixel values ​​in each sector area is used as the pixel statistical feature of the image, and the first target monitoring image feature is generated based on the variance of all pixel values ​​in each sector area and the image color component feature; The method performs key point detection on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain corresponding target key points, and generates a second target monitoring image feature based on the H channel image, the S channel image and the V channel image in combination with the target key points, including: constructing a Gaussian pyramid based on a Gaussian function, generating a differential Gaussian scale space based on the Gaussian pyramid using a differential Gaussian scale space function, obtaining differences between adjacent scale layers in the Gaussian scale space based on the Gaussian pyramid, and generating a differential Gaussian scale space based on differences between adjacent scale layers in combination with the differential Gaussian scale space function; performing extreme point search processing on the preprocessed target monitoring image based on the differential Gaussian scale space to obtain a number of corresponding extreme points, and screening a number of extreme points based on a preset screening rule, that is, removing points with low contrast and unstable edge response points, and calculating an objective function value for each extreme point based on the objective function. If the objective function value does not reach a preset interval, it is determined that the contrast of the extreme point is low, and the extreme point is removed, and the principal curvature and the relative curvature of each extreme point in the direction across the edge are calculated by a Gaussian difference operator. The principal curvature in the vertical edge direction is compared with the principal curvature in the cross-edge direction and the principal curvature in the vertical edge direction. If the principal curvature in the cross-edge direction is greater than the principal curvature in the vertical edge direction, the extreme point is judged to be an unstable edge response point, and the extreme point is removed until all points with low contrast and unstable edge response points are removed. The retained extreme points are obtained, that is, the screened extreme points, and the screened extreme points are used as target key points corresponding to the preprocessed target monitoring image; a neighborhood window is set based on the target key point, the color mean of all pixel points of the H channel image, the S channel image and the V channel image in the neighborhood window is calculated, and color features are generated based on the color mean of all pixel points in the neighborhood window; the H channel image, the S channel image and the V channel image are divided into regions based on the target key point to obtain several pixel regions, and the corresponding directional gradient histogram is extracted for each pixel region; the second target monitoring image feature is generated based on the directional gradient histogram of all pixel regions and the color feature.

7. An electronic device, comprising a processor and a memory, characterized in that: The memory is used to store instructions, and the processor is used to call the instructions in the memory, so that the electronic device executes the machine vision-based aquatic area monitoring and analysis method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed on an electronic device, the electronic device executes the machine vision-based aquatic area monitoring and analysis method as described in any one of claims 1 to 5.

Images