A Smart Detection Method and System for Microplastic Pollutants Based on Image Recognition

Abstract

The present invention discloses an intelligent detection method and system for microplastic pollutants based on image recognition, which belongs to the field of image recognition technology, and specifically comprises: setting a detection environment, sequentially putting different numbers of microplastic particles into a measured water body, and continuously collecting water body images after each putting; obtaining a grayscale image of the water body, obtaining the number of grayscale steps and the number of pixels of each grayscale step according to a grayscale histogram, calculating the grayscale sum, and obtaining the grayscale sum corresponding to each microplastic particle order of magnitude; establishing a fitting function for fitting the relationship between the grayscale sum and the change of the microplastic particle order of magnitude; obtaining a microplastic-contaminated water body to be tested, collecting the grayscale sum of the contaminated water body through image recognition, substituting the grayscale sum into the fitting function, and obtaining the corresponding microplastic particle order of magnitude; the present invention realizes efficient detection of the microplastic order of magnitude in the contaminated water body.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to an intelligent detection method and system for microplastic pollutants based on image recognition. Background Art

[0002] With the rapid development of industrialization and urbanization, plastic products are widely used in various fields, resulting in the generation of a large amount of plastic waste. These plastic wastes are difficult to degrade in the natural environment, forming microplastic pollutants, which pose a serious threat to the environment and ecosystem. Microplastic pollutants refer to plastic particles with a diameter of less than 5mm. They are widely present in water, soil and atmosphere, posing potential risks to biodiversity and human health.

[0003] At present, the detection of microplastic pollutants mainly relies on sampling and laboratory analysis methods. These methods usually require time-consuming sample collection, pretreatment and instrumental analysis processes, making it difficult to achieve rapid and real-time monitoring. In addition, existing detection methods often rely on the operation and judgment of professionals, which have certain subjectivity and errors, and due to the influence of the dispersion state of microplastics, it is difficult to achieve a rapid judgment of the order of magnitude of microplastics. Therefore, it is of great significance to develop an efficient and accurate intelligent detection method for microplastic pollutants. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent detection method and system for microplastic pollutants based on image recognition to solve the following technical problems:

[0005] The detection of microplastic pollutants mainly relies on sampling and laboratory analysis methods. Due to the influence of the dispersion state of microplastics, it is difficult to quickly judge the order of magnitude of microplastics.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] An intelligent detection method for microplastic pollutants based on image recognition comprises the following steps:

[0008] Setting a detection environment, wherein the detection environment includes a detection pool of fixed volume, irradiating the detection pool with light of fixed incident angle, injecting a preset volume of measurement water into the detection pool, and sequentially placing different amounts of microplastic particles into the measurement water, until the microplastic particles are evenly dispersed after each placement, and continuously collecting water images;

[0009] The water body image is subjected to noise reduction and grayscale processing to obtain a grayscale image, a grayscale histogram of the grayscale image is generated, the number of grayscale steps n and the number of pixels mi of each grayscale step Gi are obtained according to the grayscale histogram, i=1,...,n, and the total grayscale value is calculated. , obtain the grayscale sum corresponding to each microplastic particle magnitude N;

[0010] A fitting function is established based on the power function, and the fitting function is used to fit the relationship between the total grayscale and the change of the order of magnitude N of microplastic particles; a microplastic-contaminated water body to be tested is obtained, and the contaminated water body of the same volume as the measured water body is extracted and placed in the detection environment, and the total grayscale of the contaminated water body is collected through image recognition, and the total grayscale is substituted into the fitting function to obtain the corresponding order of magnitude of microplastic particles, and the output order of magnitude is calibrated based on the image recognition result.

[0011] As a further solution of the present invention, the specific process of denoising the water body image is as follows:

[0012] Based on the bilateral filter function of OpenCV, the water image is denoised. The pixel value weight and spatial distance weight are set to perform denoising on the flat area and edge area of ​​the water image respectively.

[0013] As a further solution of the present invention: the formula of the power function is:

[0014] ;

[0015] Wherein exp represents a natural exponential function, a and b are parameters, and a is greater than 0, and b is less than 0; the power exponential function is transformed into a linear regression model Y=α+βX, wherein Y=ln[ln(1 / Sum)], X=ln(N), α=ln(-a), β=b, and the values ​​of α and β are calculated by the least squares method, and the values ​​of parameters a and b are obtained.

[0016] As a further solution of the present invention: the specific definition of the grayscale sum corresponding to each microplastic particle order N is:

[0017] After each release, until the microplastic particles are evenly dispersed, water images are continuously collected, and the grayscale sums corresponding to the water images under different dispersion states are calculated respectively. The average grayscale sums are used to establish the fitting function.

[0018] As a further solution of the present invention: the process of calibrating the output magnitude is:

[0019] Obtain images of any dispersed state of all microplastic particles, calculate the vector A corresponding to each image through the feature representation f(θ), and calculate the average vector A corresponding to each dispersed state. j , j = 1, ..., M, where M represents the number of all scattered states of all orders of magnitude, and the average vector A j The labels are the vector representations corresponding to the scattered states;

[0020] For the collected polluted water body image, the vector representation of the polluted water body image is calculated through the feature representation f(θ), marked as B, and the similarity p between the polluted water body and each scattered state is calculated. The formula is;

[0021] ;

[0022] Where x represents the polluted water image, y represents the dispersion state, and τ represents the similarity scaling factor. j , B> means A j and the cosine distance of A and B; according to the similarity output result, obtain the similar dispersion state with the highest similarity to the polluted water body, and the similar order of magnitude to which the similar dispersion state belongs. When the order of magnitude output by the fitting function is consistent with the similar order of magnitude, the order of magnitude result is directly output; when it does not meet the requirements, the difference between the grayscale sum of the similar dispersion state and the average grayscale sum is obtained. If the difference is greater than one order of magnitude, the order of magnitude result is still output, otherwise the polluted water body image is re-collected.

[0023] As a further solution of the present invention: when the order of magnitude output by the fitting function still does not match the similarity order of magnitude after the polluted water body image is re-collected, the dispersion state of the polluted water body is adjusted.

[0024] The present invention also includes an intelligent detection system for microplastic pollutants based on image recognition, comprising:

[0025] An image acquisition module is used to acquire water body images in a detection environment, wherein the detection environment includes a detection pool of fixed volume, and irradiates the detection pool with light of a fixed incident angle, injects a preset volume of measurement water into the detection pool, and sequentially releases different amounts of microplastic particles into the measurement water, and continuously acquires water body images after each release until the microplastic particles are evenly dispersed;

[0026] An image processing model is used to perform noise reduction and grayscale processing on a water body image to obtain a grayscale image, generate a grayscale histogram of the grayscale image, obtain the number of grayscale steps n and the number of pixels mi of each grayscale step Gi according to the grayscale histogram, i=1,...,n, and calculate the total grayscale value. , obtain the grayscale sum corresponding to each microplastic particle magnitude N;

[0027] The result output module is used to establish a fitting function based on the power function, and the fitting function is used to fit the relationship between the total grayscale and the change of the order of magnitude N of microplastic particles; obtain the microplastic contaminated water body to be tested, extract the contaminated water body of the same volume as the measured water body and put it into the detection environment, collect the total grayscale of the contaminated water body through image recognition, substitute the total grayscale into the fitting function, obtain the corresponding order of magnitude of microplastic particles, and calibrate the output order of magnitude based on the image recognition result.​

[0028] Beneficial effects of the present invention:

[0029] (1) The present invention can detect microplastic pollutants with high precision by setting a specific detection environment and adopting image recognition technology. It uses computer vision and image processing technology to perform noise reduction, grayscale processing and feature extraction on water images, realizes an automated processing process, accurately identifies and counts microplastic particles, and improves the accuracy and reliability of detection. When the number of microplastic particles in the water body gradually increases, the reflectivity of the water body will gradually decrease, and the reflectivity is directly reflected in the change of grayscale. Therefore, the present invention represents the reflectivity by the sum of grayscales. Based on the change of the sum of grayscales with the order of magnitude of microplastic particles, a fitting function is established, which can fit the relationship between the sum of grayscales and the change of the order of magnitude of microplastic particles. This method can more accurately describe the relationship between the number of microplastic particles and the sum of grayscales, thereby improving the accuracy of the detection results.

[0030] (2) The present invention further improves the accuracy and reliability of the detection results by calculating the similarity between the polluted water body and each dispersion state and calibrating according to the similarity order of magnitude of the similar dispersion state with the highest similarity. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The present invention will be further described below in conjunction with the accompanying drawings.

[0032] Figure 1 It is a schematic diagram of the process of an intelligent detection method for microplastic pollutants based on image recognition of the present invention. DETAILED DESCRIPTION

[0033] 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.

[0034] See also Figure 1 As shown, the present invention is an intelligent detection method for microplastic pollutants based on image recognition, comprising the following steps:

[0035] 1. Set up the test environment: First, build a standardized test environment consisting of a test pool with a fixed volume. In the test pool, irradiate light at a fixed incident angle to ensure the consistency of lighting conditions. Then, inject a preset volume of measurement water into the test pool to provide a standardized water sample basis for subsequent experiments.

[0036] 2. Microplastic particle placement and image acquisition: Then, different orders of magnitude of microplastic particles are placed in the measured water body in turn. After each placement, it is necessary to wait until the microplastic particles are evenly dispersed in the water. This process may involve a certain amount of stirring or standing time to ensure the uniformity of particle distribution. Subsequently, high-resolution camera equipment is used to continuously collect water body images to record the distribution of microplastic particles in the water body at each order of magnitude.

[0037] 3. Image preprocessing: The collected water body images are subjected to noise reduction processing to reduce the impact of environmental noise and equipment errors on image quality. Subsequently, the color images are converted to grayscale images for subsequent analysis. By generating a grayscale histogram of the grayscale image, the distribution of each grayscale level in the image can be understood in detail. According to the grayscale histogram, the number of grayscale steps and the number of pixels in each grayscale step are determined. This information is crucial for subsequent data analysis.

[0038] 4. Calculate the total grayscale: Based on the determined grayscale levels and the number of pixels, calculate the total grayscale of each grayscale level. This step is the basis for establishing the relationship between the magnitude of microplastic particles and the total grayscale. By comparing the total grayscale at different magnitudes, we can preliminarily explore the change pattern between the two.

[0039] 5. Establish a fitting function: In order to more accurately describe the relationship between the total grayscale and the change in the order of magnitude N of microplastic particles, the power function is used to establish the fitting function. The power function performs well in describing complex nonlinear relationships due to its flexibility and wide adaptability. The specific parameters in the power function are determined by parameter estimation methods such as the least squares method, thereby obtaining an accurate fitting model.

[0040] 6. Detection of polluted water bodies to be tested: When the water bodies to be tested are contaminated by microplastics, first extract a sample of the same volume as the previously measured water body and put it into the test environment. Then, the grayscale sum of the polluted water body is collected through image recognition technology. Substituting this grayscale sum into the previously established fitting function, the corresponding order of magnitude of microplastic particles can be calculated. However, due to the complexity and uncertainty of the actual water environment, the order of magnitude of the direct output may need further calibration.

[0041] 7. Output result calibration: Based on the image recognition results, calibrate the magnitude of the fitting function output. This involves comparing with other known data or standards, or optimizing with more complex algorithms. Calibration can improve the accuracy and reliability of the test results, providing strong support for the monitoring and governance of microplastic pollution.

[0042] In another preferred embodiment of the present invention, the specific process of denoising the water body image is as follows:

[0043] Based on OpenCV's bilateral filter function, bilateralFilter is used to reduce noise in water images. The bilateralFilter function controls the degree of noise reduction by setting pixel value weights and spatial distance weights. For flat areas of water images, since the pixel values ​​in these areas vary little, a larger pixel value weight can be set to enhance the smoothing effect. For edge areas of water images, since the pixel values ​​in these areas vary greatly, a smaller pixel value weight needs to be set to avoid over-smoothing and resulting in blurred edges.

[0044] In addition, the bilateralFilter function can also control the range of noise reduction by setting the spatial distance weight. For the flat area of ​​the water image, a larger spatial distance weight can be set to expand the smoothing range and reduce noise. For the edge area of ​​the water image, a smaller spatial distance weight needs to be set to maintain the clarity and details of the edge.

[0045] By reasonably setting the pixel value weight and spatial distance weight, the bilateralFilter function can effectively perform noise reduction on different water image areas.

[0046] In another preferred embodiment of the present invention, the formula of the power function is:

[0047] ;

[0048] Wherein exp represents a natural exponential function, a and b are parameters, and a is greater than 0, and b is less than 0; the power exponential function is transformed into a linear regression model Y=α+βX, wherein Y=ln[ln(1 / Sum)], X=ln(N), α=ln(-a), β=b, and the values ​​of α and β are calculated by the least squares method, and the values ​​of parameters a and b are obtained.

[0049] The reflectivity (i.e., reflectivity) of the water surface usually does not decrease linearly with the increase in the concentration of microplastic particles. At low concentrations, microplastic particles may have little effect on the reflectivity, but as the concentration increases, the particles may significantly change the optical properties of the water surface. Therefore, the present invention represents the reflectivity by the sum of grayscale, establishes a fitting function based on the power function, and can fit the relationship between the sum of grayscale and the change in the order of magnitude of microplastic particles according to the characteristic that the rate of decrease of the power function image increases with the increase of the horizontal coordinate. This method can more accurately describe the relationship between the number of microplastic particles and the sum of grayscale, thereby improving the accuracy of the detection results.

[0050] In another preferred embodiment of the present invention, the grayscale sum corresponding to each microplastic particle order N is specifically defined as:

[0051] After each release of microplastic particles, it is necessary to wait for the particles to be evenly dispersed in the water. In this process, it is crucial to continuously collect images of the water body. Through these images, the morphology and distribution of microplastic particles in different dispersion states can be captured. Subsequently, the sum of the grayscale of the images in each dispersion state is calculated separately, which is an important basis for evaluating the number of microplastic particles. Finally, the sum of the grayscale in each dispersion state is averaged to establish a fitting function to more accurately describe the relationship between the number of microplastic particles and the sum of the grayscale.

[0052] Furthermore, the process of calibrating the output magnitude is:

[0053] 1. Obtain images of dispersed states: First, select and obtain images of any dispersed state at any order of magnitude from all known microplastic particle orders of magnitude. These images should cover the distribution of microplastic particles in water bodies at different orders of magnitude to provide comprehensive reference data.

[0054] 2. Calculate vector representation: For each scattered image, calculate its corresponding vector A through the feature extraction function f(θ). Here, the feature extraction function f(θ) can be any algorithm that can effectively describe the characteristics of the image, such as SIFT, SURF, etc. Similarly, for the collected polluted water image, its vector representation B is also calculated through the same feature extraction function f(θ). This step aims to convert the image into a quantifiable mathematical representation to facilitate subsequent similarity calculation and analysis.

[0055] 3. Calculate the average vector: For each dispersion state, calculate the average vector A of all images in that state j By calculating the average vector, we can eliminate the error or noise that may be caused by a single image and improve the stability and accuracy of the model. j The labels are the vector representations of the corresponding scattered states.

[0056] 4. Calculate similarity: Next, for the collected polluted water image, the vector representation B calculated by the feature extraction function f(θ) is compared with the average vector A of each scattered state. j The similarity p is calculated using the formula:

[0057] ;

[0058] Where x represents the polluted water image, y represents the dispersion state, and τ represents the similarity scaling factor. j , B> means A j The cosine distance from A to B.

[0059] ​5. Determine the most similar dispersion state: Based on the calculated similarity value, determine the dispersion state with the highest similarity to the polluted water body. This step is the core of the entire calibration process because it directly determines the accuracy of the order of magnitude results of the final output. After finding the most similar dispersion state, further determine the similarity order of magnitude to which the dispersion state belongs.

[0060] 6. Compare the output of the fitting function with the similar order of magnitude: Compare the order of magnitude of the output of the fitting function with the determined similar order of magnitude. If the two match, it means that the fitting function has good applicability in the current situation, and the result of the fitting function can be directly output as the final order of magnitude. However, if the two do not match, further processing is required.

[0061] 7. Handling discrepancies: When the order of magnitude output by the fitting function does not match the similar order of magnitude, first calculate the difference between the sum of the grayscale of the similar dispersion state and the average grayscale sum. This difference reflects the degree of deviation between the current dispersion state and the average level. If the difference is greater than a set threshold (i.e., an order of magnitude), the current polluted water condition is considered to be special, and the result of the fitting function is still output as the order of magnitude. However, if the difference is less than the threshold, it means that the current polluted water condition is significantly different from the expected condition, and the polluted water image needs to be re-collected and calibrated.

[0062] It is worth noting that after re-collecting the polluted water image, if the order of magnitude of the fitting function output is still inconsistent with the similar order of magnitude, it indicates that the current dispersion state of the polluted water may be abnormal or erroneous. At this time, the dispersion state of the polluted water needs to be adjusted, such as by changing the stirring speed, standing time or sampling position, to obtain a more uniform and accurate dispersion state. It can more accurately simulate the actual situation of the polluted water, so that the output of the fitting function is closer to the true value, improving the accuracy and reliability of the detection.

[0063] The present invention also includes an intelligent detection system for microplastic pollutants based on image recognition, comprising:

[0064] An image acquisition module is used to acquire water body images in a detection environment, wherein the detection environment includes a detection pool of fixed volume, and irradiates the detection pool with light of a fixed incident angle, injects a preset volume of measurement water into the detection pool, and sequentially releases different amounts of microplastic particles into the measurement water, and continuously acquires water body images after each release until the microplastic particles are evenly dispersed;

[0065] An image processing model is used to perform noise reduction and grayscale processing on a water body image to obtain a grayscale image, generate a grayscale histogram of the grayscale image, obtain the number of grayscale steps n and the number of pixels mi of each grayscale step Gi according to the grayscale histogram, i=1,...,n, and calculate the total grayscale value. , obtain the grayscale sum corresponding to each microplastic particle magnitude N;

[0066] The result output module is used to establish a fitting function based on the power function, and the fitting function is used to fit the relationship between the total grayscale and the change of the order of magnitude N of microplastic particles; obtain the microplastic contaminated water body to be tested, extract the contaminated water body of the same volume as the measured water body and put it into the detection environment, collect the total grayscale of the contaminated water body through image recognition, substitute the total grayscale into the fitting function, obtain the corresponding order of magnitude of microplastic particles, and calibrate the output order of magnitude based on the image recognition result.

[0067] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. An intelligent detection method for microplastic pollutants based on image recognition, characterized in that: The following steps are involved: Setting a detection environment, wherein the detection environment includes a detection pool of fixed volume, irradiating the detection pool with light of fixed incident angle, injecting a preset volume of measurement water into the detection pool, and sequentially placing different amounts of microplastic particles into the measurement water, until the microplastic particles are evenly dispersed after each placement, and continuously collecting water images; The water body image is subjected to noise reduction and grayscale processing to obtain a grayscale image, a grayscale histogram of the grayscale image is generated, the number of grayscale steps n and the number of pixels mi of each grayscale step Gi are obtained according to the grayscale histogram, i=1,...,n, and the total grayscale value is calculated. , obtain the grayscale sum corresponding to each microplastic particle magnitude N; A fitting function is established based on the power function, and the fitting function is used to fit the relationship between the total grayscale and the change of the order of magnitude N of microplastic particles; a microplastic-contaminated water body to be tested is obtained, and the contaminated water body of the same volume as the measured water body is extracted and placed in the detection environment, and the total grayscale of the contaminated water body is collected through image recognition, and the total grayscale is substituted into the fitting function to obtain the corresponding order of magnitude of microplastic particles, and the output order of magnitude is calibrated based on the image recognition result.

2. According to claim 1, a method for intelligent detection of microplastic pollutants based on image recognition is characterized in that: The specific process of denoising water images is as follows: Based on the bilateral filter function of OpenCV, the water image is denoised. The pixel value weight and spatial distance weight are set to perform denoising on the flat area and edge area of ​​the water image respectively.

3. According to claim 1, a method for intelligent detection of microplastic pollutants based on image recognition is characterized in that: The formula of the power function is: ； Wherein exp represents a natural exponential function, a and b are parameters, and a is greater than 0, and b is less than 0; the power exponential function is transformed into a linear regression model Y=α+βX, wherein Y=ln[ln(1 / Sum)], X=ln(N), α=ln(-a), β=b, and the values ​​of α and β are calculated by the least squares method, and the values ​​of parameters a and b are obtained.

4. According to claim 1, a method for intelligent detection of microplastic pollutants based on image recognition is characterized in that: The specific definition of the total grayscale corresponding to each microplastic particle magnitude N is: After each release, until the microplastic particles are evenly dispersed, water images are continuously collected, and the grayscale sums corresponding to the water images under different dispersion states are calculated respectively. The average grayscale sums are used to establish the fitting function.

5. According to claim 4, a method for intelligent detection of microplastic pollutants based on image recognition is characterized in that: The process of calibrating the output magnitude is: Obtain images of any dispersed state of all microplastic particles, calculate the vector A corresponding to each image through the feature representation f(θ), and calculate the average vector A corresponding to each dispersed state. j , j = 1, ..., M, where M represents the number of all scattered states of all orders of magnitude, and the average vector A j The labels are the vector representations corresponding to the scattered states; For the collected polluted water image, the vector representation of the polluted water image is calculated through the feature representation f(θ), marked as B, and the similarity p between the polluted water and each scattered state is calculated. The formula is: ； Where x represents the polluted water image, y represents the dispersion state, and τ represents the similarity scaling factor. j , B> means A j and the cosine distance of A and B; according to the similarity output result, obtain the similar dispersion state with the highest similarity to the polluted water body, and the similar order of magnitude to which the similar dispersion state belongs. When the order of magnitude output by the fitting function is consistent with the similar order of magnitude, the order of magnitude result is directly output; when it does not meet the requirements, the difference between the grayscale sum of the similar dispersion state and the average grayscale sum is obtained. If the difference is greater than one order of magnitude, the order of magnitude result is still output, otherwise the polluted water body image is re-collected.​ 6. The method for intelligent detection of microplastic pollutants based on image recognition according to claim 5 is characterized in that: When the order of magnitude output by the fitting function still does not match the similar order of magnitude after the polluted water body image is re-collected, the dispersion state of the polluted water body is adjusted.

7. An intelligent detection system for microplastic pollutants based on image recognition, characterized in that: include: An image acquisition module is used to acquire water body images in a detection environment, wherein the detection environment includes a detection pool of fixed volume, and irradiates the detection pool with light of a fixed incident angle, injects a preset volume of measurement water into the detection pool, and sequentially releases different amounts of microplastic particles into the measurement water, and continuously acquires water body images after each release until the microplastic particles are evenly dispersed; An image processing model is used to perform noise reduction and grayscale processing on a water body image to obtain a grayscale image, generate a grayscale histogram of the grayscale image, obtain the number of grayscale steps n and the number of pixels mi of each grayscale step Gi according to the grayscale histogram, i=1,...,n, and calculate the total grayscale value. , obtain the grayscale sum corresponding to each microplastic particle magnitude N; The result output module is used to establish a fitting function based on the power function, and the fitting function is used to fit the relationship between the total grayscale and the change of the order of magnitude N of microplastic particles; obtain the microplastic contaminated water body to be tested, extract the contaminated water body of the same volume as the measured water body and put it into the detection environment, collect the total grayscale of the contaminated water body through image recognition, substitute the total grayscale into the fitting function, obtain the corresponding order of magnitude of microplastic particles, and calibrate the output order of magnitude based on the image recognition result.

Images