Algae control system and method based on machine learning algorithm and image recognition

Abstract

The present application provides an algae control system and method based on machine learning algorithms and image recognition, which belongs to the technical field of water algae control and prevention. The present application constructs a chlorophyll a concentration prediction model and an intelligent dosing model based on the LightGBM algorithm, collects real-time images of the underwater surroundings of the algae-to-be-controlled water area, and uses the Sobel operator to determine the image boundary, and determines the algae characteristic parameters through the boundary. Then, the algae characteristic parameters are input into the chlorophyll a concentration prediction model, and the obtained chlorophyll a concentration and algae characteristic parameters are input into the intelligent dosing model to obtain the algae control agent metering, and the algae control agent is sprayed according to the algae control agent metering. In addition, the present application also performs post-event monitoring of the water area after the agent is sprayed. Improve the accuracy of the algae control agent metering, achieve accurate algae control, reduce the cost of algae control, and avoid the side effects of excessive agents.

Description

Technical Field

[0001] The present invention relates to the technical field of algae control in water bodies, and in particular to an algae control system and method based on machine learning algorithms and image recognition. Background Art

[0002] Large water bodies are open and seasonally polluted, and are prone to problems such as excessive algae growth and algal blooms in summer. When the nitrogen and phosphorus content in large water bodies increases or is polluted by exogenous algae, the water bodies are prone to the risk of algal blooms. Excessive algae growth and algal blooms not only reduce the sensory pleasure of large water bodies, but may also release algal toxins into the water, causing ecological and health problems. Most of the existing algae prevention and control methods, such as enhanced coagulation and sedimentation, addition of chemical agents, ultraviolet irradiation, etc., use relatively extensive treatment methods, which have the problem of excessive energy consumption of agents and are difficult to achieve precise algae removal. Excessive agents will have side effects and affect the ecology and water quality.

[0003] Therefore, there is an urgent need for an algae control system and method for precise algae removal. Summary of the invention

[0004] In view of this, the present invention provides an algae control method based on machine learning algorithm and image recognition, characterized in that the method comprises the following steps:

[0005] S1: Construct a chlorophyll a concentration prediction model based on the LightGBM algorithm;

[0006] Build an intelligent dosing model based on the LightGBM algorithm;

[0007] S2: Collect real-time images of the underwater surrounding environment of the algae-controlled water area;

[0008] S3: determining algae characteristic parameters of the real-time image, wherein the algae characteristic parameters include algae equivalent area and algae fractal dimension;

[0009] S4: Substituting the algae characteristic parameters into the chlorophyll a concentration prediction model to obtain the chlorophyll a concentration;

[0010] S5: Determine whether the chlorophyll a concentration is greater than a preset chlorophyll a threshold value, if yes, proceed to the next step, if no, proceed to step S2;

[0011] S6: Go to the target algae area. After arriving at the target algae area, obtain the actual chlorophyll a concentration through the chlorophyll a detector, and determine whether the actual chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, proceed to the next step; if not, proceed to step S2;

[0012] S7: inputting the algae characteristic parameters and the actual chlorophyll a concentration into the intelligent dosing model to obtain the algae control agent dosage;

[0013] S8: Spray the measured amount of the algae control agent on the target algae area.

[0014] Furthermore, the method further comprises: repeating the following steps until completion;

[0015] After spraying the algae control agent, the actual chlorophyll a concentration is obtained by a chlorophyll a detector within a preset time period, and it is determined whether the actual chlorophyll a concentration is greater than a preset chlorophyll a threshold value. If so, repeat steps S7 and S8. If not, end.

[0016] At the same time, the total amount of algae control agent dosage is determined based on the obtained algae control agent dosage.

[0017] Furthermore, the method also includes: in the process of going to the target algae area, real-time detection of the chlorophyll a concentration of the passing water area, if the concentration is greater than a preset chlorophyll a threshold, pausing to go to the target algae area, obtaining the algae control agent dosage through the intelligent dosing model, spraying the algae control agent on site, and continuing to go to the target algae area after the algae control is completed.

[0018] Further, the method further comprises: step S2 comprises the following steps:

[0019] S21: Based on the initial device of the hull, adjust the light source parameters to the preset light source parameters, traverse the preset depth range from 0 to 360 degrees, and collect real-time images of the underwater surrounding environment;

[0020] S22: Determine whether the real-time image is a grayscale image. If so, proceed to step S3. If not, convert the collected underwater image into a grayscale image, wherein the grayscale value i is determined in the following manner:

[0021] i=0.3r+0.59g+0.11b (1)

[0022] Among them, i represents the grayscale value, r represents the red in the image, g represents the green in the image, and b represents the blue in the image.

[0023] Furthermore, the preset light source parameters are as follows:

[0024] The image entropy of the real-time image containing the target algae is determined, wherein the image entropy is determined by the following method:

[0025]

[0026] Among them, H represents the image entropy, P i Represents the frequency of gray value i appearing in all pixels;

[0027] Determine the light source parameters corresponding to the real-time image with the maximum image entropy;

[0028] The light source parameters corresponding to the image with the largest image entropy are preset light source parameters.

[0029] Further, step S3 includes the following steps:

[0030] S31: performing a planar convolution on the received grayscale image and the Sobel operator to obtain an algae enhanced image;

[0031] S32: binarizing the algae enhanced image using a global threshold method to obtain a dividing boundary of stacked algae;

[0032] S33: determining the algae equivalent area and fractal dimension according to the dividing boundary;

[0033] The algae equivalent area is the sum of the areas of the black areas in the binarized algae enhanced image;

[0034] The fractal dimension is determined by the following method:

[0035] ln A=D f ·ln L+lnα (3)

[0036] Where A is the projected area of ​​the algae, L is the maximum length of the projection, α is the proportional constant, and D f Represents the fractal dimension of algae in two-dimensional space.

[0037] Accordingly, the present application also provides an algae control system based on machine learning algorithm and image recognition, the system is applicable to an algae control method based on machine learning algorithm and image recognition, and is characterized in that it includes:

[0038] Image acquisition module: used to obtain real-time images of the underwater environment of the algae area to be controlled;

[0039] An image analysis module: used to determine the stacking boundary of the target algae in the underwater image, and determine the algae characteristic parameters of the underwater image according to the stacking boundary;

[0040] Chlorophyll a concentration prediction module: used to predict the chlorophyll a concentration corresponding to the algae characteristic parameters according to the algae characteristic parameters;

[0041] Intelligent decision-making module: used to determine whether the chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, the power module is controlled to go to the target algae area, and the algae control agent dosage is determined according to the intelligent dosing model and output to the algae control module. If not, no processing is performed; at the same time, the intelligent decision-making module is used to perform post-algae control monitoring for each target algae area;

[0042] Power module: used to go to the target algae area according to the instructions of the intelligent decision-making module;

[0043] GPS module: used to obtain the location of the algae control system;

[0044] Chlorophyll a detector: used to obtain the chlorophyll a concentration in the target algae area after reaching the area;

[0045] Algae control module: spray the algae control agent measured by the output of the intelligent decision-making module.

[0046] Furthermore, the image acquisition module includes a fill light device, and the fill light device is used to adjust light source parameters.

[0047] Beneficial technical effects of the present invention: the algae control method of the present invention uses the Sobel algorithm to accurately divide the edges of the algae picture and determine the algae characteristic parameters, and uses the LightGBM algorithm to determine the chlorophyll a concentration with the algae characteristic parameters as input, so as to quickly find the areas with excessive algae in the water body, and then construct a dosing model based on the LightGBM algorithm, and input the chlorophyll a concentration and the algae characteristic parameters into the dosing model to obtain the algae control agent dosage, thereby achieving precise algae control and reducing the cost of algae control; the algae control dosage is more accurate and efficient, and by re-examining the historical algae outbreak sites, the risk of algae outbreaks is further reduced, thereby achieving long-term suppression of water blooms. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0049] Figure 1 This is a flow chart of the algae control method of this application. DETAILED DESCRIPTION

[0050] The present invention is further described below in conjunction with the accompanying drawings:

[0051] The present invention provides an algae control method based on machine learning algorithm and image recognition, characterized in that the method comprises the following steps:

[0052] S1: Construct a chlorophyll a concentration prediction model based on the LightGBM algorithm; establish a chlorophyll a concentration prediction model with the algae characteristic parameters as input and the corresponding chlorophyll a concentration as output through the LightGBM machine learning algorithm, and determine whether the target algae area has the risk of algal bloom outbreak through the chlorophyll a concentration prediction model.

[0053] An intelligent dosing model is constructed based on the LightGBM algorithm; through the LightGBM machine learning algorithm, an intelligent dosing model is established with the algae characteristic parameters of the target algae, the corresponding chlorophyll a concentration, and the algae control dosage plan as input, and the chlorophyll a concentration after algae control as output.

[0054] The input of the dosing model constructed in this step is: algae characteristic parameters, chlorophyll a concentration and algae control agent dosage, and the output is: chlorophyll a concentration after algae control. The LightGBM algorithm searches for the optimal dose within a given algae control dose range, that is, it traverses the possible doses (or combinations) within the dose range, and selects the optimal dose (or combination) according to the target value of chlorophyll a concentration after algae control.

[0055] The LightGBM model described in this embodiment is implemented in Python language, and the dosing model and chlorophyll a concentration prediction model based on the LightGBM algorithm are both implemented using the lightgbm machine learning algorithm library.

[0056] S2: Collect real-time images of the underwater surrounding environment of the algae-controlled water area;

[0057] S21: Based on the initial device of the hull, adjust the light source parameters to the preset light source parameters, traverse the preset depth range of 0 to 360 degrees, and collect real-time images of the underwater surrounding environment; in this embodiment, the real-time image collection device is a waterproof industrial camera, whose lens is a zoom lens with a maximum optical zoom of 20 times, and the pixel of the industrial camera is greater than or equal to 2 million, and the frame rate is 60 frames per second. Based on the initial position of the waterproof industrial camera, rotate clockwise from 0 to 360 degrees, and search for the target algae in the vertical direction of 0.25 / 0.5 / 1 / 2 / 3 / 4 / 5 meters underwater as the criterion, and any object may be identified as the target algae.

[0058] The waterproof industrial camera collects the target algae image in real time, and the input image is the video image collected by the industrial camera. The original image read from the video stream is converted into a grayscale image.

[0059] S22: Determine whether the real-time image is a grayscale image. If so, proceed to step S3. If not, convert the collected underwater image into a grayscale image, wherein the grayscale value i is determined in the following manner:

[0060] i=0.3r+0.59g+0.11b (1)

[0061] Among them, i represents the grayscale value, r represents the red in the image, g represents the green in the image, and b represents the blue in the image.

[0062] S3: Determine the stacking boundary of the target algae in the real-time image, and determine the algae characteristic parameters of the real-time image according to the stacking boundary, the algae characteristic parameters include algae equivalent area, algae equivalent perimeter, algae fractal dimension and algae texture characteristics; the image analysis module can perform edge division of a single algae community on the target algae image through operators such as Sobel. The target algae means that any object in the underwater real-time image may be identified as the target algae. In this embodiment, when the chlorophyll a predicted value of the algae area to be identified is greater than the set value, the algae area is considered to be the target algae area.

[0063] S4: Substituting the algae characteristic parameters into the chlorophyll a concentration prediction model to obtain the chlorophyll a concentration;

[0064] S5: Determine whether the chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, proceed to the next step, if not, proceed to step S2; the preset chlorophyll a threshold is determined based on empirical values, and technicians in this field can set the chlorophyll a threshold based on actual algae control accuracy requirements.

[0065] S6: Go to the target algae area through the power module. After arriving at the target algae area, obtain the actual chlorophyll a concentration through the chlorophyll a detector, and determine whether the actual chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, proceed to the next step, if not, proceed to step S2; in this embodiment, the waterproof industrial camera and the sobel algorithm have real-time focus tracking function during collaborative use. When the chlorophyll a predicted value of the algae area to be identified is greater than the set value, the algae area is considered to be the target algae area. Take the center of the equivalent circular area of ​​the algae area as the forward target, adjust the forward direction of the hull, and in the process of advancing, accurately calculate the center position of the equivalent circular area of ​​the target algae area in real time and adjust the direction in time until it arrives. Among them, the center position of the equivalent circular area of ​​the target algae area is determined by the equivalent area of ​​the target algae, and the center position is the geometric center of the equivalent area.

[0066] S7: inputting the algae characteristic parameters and the actual chlorophyll a concentration into the intelligent dosing model to obtain the algae control agent dosage;

[0067] S8: Spray the measured amount of the algae control agent on the target algae area.

[0068] In this embodiment, the method further includes: repeating the following steps until completion;

[0069] After spraying the algae control agent, the actual chlorophyll a concentration is obtained by a chlorophyll a detector within a preset time period, and it is determined whether the actual chlorophyll a concentration is greater than a preset chlorophyll a threshold value. If so, repeat steps S7 and S8. If not, end.

[0070] At the same time, the total amount of algae control agent dosage is determined based on the obtained algae control agent dosage.

[0071] In this embodiment, the target algae area after the algae control treatment will be tracked and monitored to determine whether to perform multiple algae control treatments. The preset time period can be set by those skilled in the art according to actual needs. In this embodiment, the preset time period is between 24 hours and 30 hours after each algae control treatment. The chlorophyll a concentration of the target algae area is obtained by a chlorophyll a detector. If it is lower than the set value, no algae control treatment will be performed. Otherwise, the algae control dosage plan is given according to the real-time chlorophyll a concentration and the algae characteristic parameters through the intelligent dosing model of the LightGBM machine learning algorithm. At the same time, the intelligent dosing model will update the algae control dosage plan to the sum of multiple algae control dosages. The sum data of the algae control agent is used to update the intelligent dosing model. By reviewing the historical algae outbreak sites, the risk of algae outbreaks is further reduced, and long-term suppression of water blooms is achieved.

[0072] In this embodiment, the method further includes: in the process of going to the target algae area, the chlorophyll a concentration of the passing water area is detected in real time. If the concentration is greater than the preset chlorophyll a threshold, the process of going to the target algae area is suspended, and the algae control agent dosage is obtained through the intelligent dosing model, and the algae control agent dosage is sprayed on the spot, and after the algae control is completed, the process of going to the target algae area is continued. The algae control effect of the algae control area to be controlled is ensured, and if an algae control area is found, it is immediately processed, and then the target algae area is traveled after the processing, which can save the time of round trip and improve the efficiency of algae control.

[0073] In this embodiment, the preset light source parameters are as follows:

[0074] The image entropy of the real-time image containing the target algae is determined, wherein the image entropy is determined by the following method:

[0075]

[0076] Among them, H represents the image entropy, which indicates the frequency of gray value i appearing in all pixels;

[0077] Determine the light source parameters corresponding to the real-time image with the maximum image entropy;

[0078] The light source parameters corresponding to the image with the maximum image entropy are preset light source parameters. In this embodiment, the image information entropy corresponding to each focal length is determined by adjusting the light source parameters, thereby determining the maximum image information entropy and the corresponding brightness.

[0079] In this embodiment, step S3 includes the following steps:

[0080] S31: performing a planar convolution on the received grayscale image and the Sobel operator to obtain an algae enhanced image;

[0081] S32: binarizing the algae enhanced image using a global threshold method to obtain a dividing boundary of stacked algae;

[0082] S33: determining the algae equivalent area and fractal dimension according to the dividing boundary;

[0083] The algae equivalent area is the sum of the areas of the black areas in the binarized algae enhanced image;

[0084] The fractal dimension is determined by the following method:

[0085] ln A=D f ·ln L+lnα (3)

[0086] Where A is the projected area of ​​the algae, L is the maximum length of the projection, α is the proportional constant, and D f Represents the fractal dimension of algae in two-dimensional space.

[0087] Accordingly, the present application also provides an algae control system based on machine learning algorithm and image recognition, the system is applicable to an algae control method based on machine learning algorithm and image recognition, and is characterized in that it includes:

[0088] Image acquisition module: used to obtain real-time images of the underwater environment of the algae area to be controlled;

[0089] An image analysis module: used to determine the stacking boundary of the target algae in the underwater image, and determine the algae characteristic parameters of the underwater image according to the stacking boundary;

[0090] Chlorophyll a concentration prediction module: used to predict the chlorophyll a concentration corresponding to the algae characteristic parameters according to the algae characteristic parameters;

[0091] Intelligent decision-making module: used to determine whether the chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, the power module is controlled to go to the target algae area, and the algae control agent dosage is determined according to the intelligent dosing model and output to the algae control module. If not, no processing is performed; at the same time, the intelligent decision-making module is used to perform post-algae control monitoring for each target algae area;

[0092] Power module: used to go to the target algae area according to the instructions of the intelligent decision-making module;

[0093] GPS module: used to obtain the location of the algae control system;

[0094] Chlorophyll a detector: used to obtain the chlorophyll a concentration in the target algae area after reaching the area;

[0095] Algae control module: spray the algae control agent measured by the output of the intelligent decision-making module.

[0096] In this embodiment, the image acquisition module includes a fill light device, and the fill light device is used to adjust the light source parameters. The fill light device includes existing lighting equipment, such as LED lights and infrared lights, which will not be described in detail here.

[0097] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A method for controlling algae based on machine learning algorithm and image recognition, characterized in that: The method comprises the following steps: S1: Construct a chlorophyll a concentration prediction model based on the LightGBM algorithm; Build an intelligent dosing model based on the LightGBM algorithm; S2: Collect real-time images of the underwater surrounding environment of the algae-controlled water area; S3: determining algae characteristic parameters of the real-time image, wherein the algae characteristic parameters include algae equivalent area and algae fractal dimension; S4: Substituting the algae characteristic parameters into the chlorophyll a concentration prediction model to obtain the chlorophyll a concentration; S5: Determine whether the chlorophyll a concentration is greater than a preset chlorophyll a threshold value, if yes, proceed to the next step, if no, proceed to step S2; S6: Go to the target algae area. After arriving at the target algae area, obtain the actual chlorophyll a concentration through the chlorophyll a detector, and determine whether the actual chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, proceed to the next step; if not, proceed to step S2; S7: inputting the algae characteristic parameters and the actual chlorophyll a concentration into the intelligent dosing model to obtain the algae control agent dosage; S8: Spray the measured amount of the algae control agent on the target algae area.

2. The algae control method based on machine learning algorithm and image recognition according to claim 1, characterized in that: The method further comprises: repeating the following steps until completion; After spraying the algae control agent, the actual chlorophyll a concentration is obtained by a chlorophyll a detector within a preset time period, and it is determined whether the actual chlorophyll a concentration is greater than a preset chlorophyll a threshold value. If so, repeat steps S7 and S8. If not, end. At the same time, the total amount of algae control agent dosage is determined based on the obtained algae control agent dosage.

3. The algae control method based on machine learning algorithm and image recognition according to claim 2, characterized in that: The method also includes: in the process of going to the target algae area, the chlorophyll a concentration of the passing water area is detected in real time. If the concentration is greater than a preset chlorophyll a threshold, the process of going to the target algae area is suspended, and the dosage of the algae control agent is obtained through the intelligent dosing model, and the dosage of the algae control agent is sprayed on the spot, and after the algae control is completed, the process of going to the target algae area continues.

4. The algae control method based on machine learning algorithm and image recognition according to claim 1, characterized in that: The method further comprises: step S2 comprises the following steps: S21: Based on the initial device of the hull, adjust the light source parameters to the preset light source parameters, traverse the preset depth range from 0 to 360 degrees, and collect real-time images of the underwater surrounding environment; S22: Determine whether the real-time image is a grayscale image. If so, proceed to step S3. If not, convert the collected underwater image into a grayscale image, wherein the grayscale value i is determined in the following manner: i=0.3r+0.59g+0.11b (1) Among them, i represents the grayscale value, r represents the red in the image, g represents the green in the image, and b represents the blue in the image.

5. The algae control method based on machine learning algorithm and image recognition according to claim 4 is characterized in that: The preset light source parameters are as follows: The image entropy of the real-time image containing the target algae is determined, wherein the image entropy is determined by the following method: Among them, H represents the image entropy, P i Represents the frequency of gray value i appearing in all pixels; Determine the light source parameters corresponding to the real-time image with the maximum image entropy; The light source parameters corresponding to the image with the largest image entropy are preset light source parameters.

6. The algae control method based on machine learning algorithm and image recognition according to claim 4, characterized in that: Step S3 includes the following steps: S31: performing a planar convolution on the received grayscale image and the Sobel operator to obtain an algae enhanced image; S32: binarizing the algae enhanced image using a global threshold method to obtain a dividing boundary of stacked algae; S33: determining the algae equivalent area and fractal dimension according to the dividing boundary; The algae equivalent area is the sum of the areas of the black areas in the binarized algae enhanced image; The fractal dimension is determined by the following method: lnA=D f ·lnL+lnα(3) Where A is the projected area of ​​the algae, L is the maximum length of the projection, α is the proportional constant, and D f Represents the fractal dimension of algae in two-dimensional space.

7. An algae control system based on machine learning algorithm and image recognition, the system is applicable to the algae control method based on machine learning algorithm and image recognition according to any one of claims 1 to 6, characterized in that: include: Image acquisition module: used to obtain real-time images of the underwater environment of the algae area to be controlled; An image analysis module: used to determine the stacking boundary of the target algae in the underwater image, and determine the algae characteristic parameters of the underwater image according to the stacking boundary; Chlorophyll a concentration prediction module: used to predict the chlorophyll a concentration corresponding to the algae characteristic parameters according to the algae characteristic parameters; Intelligent decision-making module: used to determine whether the chlorophyll a concentration is greater than the preset chlorophyll a threshold. If so, the power module is controlled to go to the target algae area, and the algae control agent dosage is determined according to the intelligent dosing model and output to the algae control module. If not, no processing is performed; at the same time, the intelligent decision-making module is used to perform post-algae control monitoring for each target algae area; Power module: used to go to the target algae area according to the instructions of the intelligent decision-making module; GPS module: used to obtain the location of the algae control system; Chlorophyll a detector: used to obtain the chlorophyll a concentration in the target algae area after reaching the area; Algae control module: spray the algae control agent measured by the output of the intelligent decision-making module.

8. The algae control system based on machine learning algorithm and image recognition according to claim 7, characterized in that: The image acquisition module includes a fill light device, and the fill light device is used to adjust light source parameters.

Images