CN114627287A - Artificial intelligence-based air-tightness detection method and system for water turbidity detection - Google Patents

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a method and a system for detecting water turbidity in air tightness detection based on artificial intelligence, wherein the method comprises the following steps: acquiring a cylinder wall image of a detection cylinder, and detecting attached bubbles on the cylinder wall to obtain an attached bubble image; clustering the attached bubbles based on the gray level to obtain bubble areas under each gray level category; selecting a sub-area with the largest attached bubble density in each bubble area; acquiring texture characteristic parameters and light and shade contrast characteristic parameters of each subregion, wherein the texture characteristic parameters comprise energy, contrast, correlation and entropy, and the light and shade contrast characteristic parameters are obtained according to the difference value of the grayscale of the subregions and the grayscale of the cylinder wall; selecting a subregion most capable of reacting the turbidity of the water by using an analytic hierarchy process; and detecting the water turbidity according to the texture characteristic parameters and the light and shade contrast characteristic parameters of the selected sub-regions. The invention can obtain more accurate and effective detection results of the water turbidity.

Description

Artificial intelligence-based air-tightness detection method and system for water turbidity detection

technical field

The invention relates to the field of artificial intelligence, in particular to a method and system for detecting turbidity of water in air tightness detection based on artificial intelligence.

Background technique

The water immersion method is one of the most commonly used traditional air tightness detection methods. However, whether it is the direct immersion method or the indirect water immersion method, the turbidity of the water body gradually increases over time. On the one hand, due to the pollutants such as oil and dust attached to the detected device, on the other hand, the water body itself carries more and more pollutants. Therefore, when the water body reaches a certain degree of turbidity, it is necessary to change the water in time to ensure the detection results. reliability. The traditional water exchange cycle basically relies on human subjective judgment, so there will be many uncertainties in the judgment of specific turbidity, which will not only waste water resources but also lead to increased errors in the detection results. The conventional judgment of turbidity is generally only based on a single factor such as brightness value and hue, and the detection result is inaccurate.

SUMMARY OF THE INVENTION

In order to solve the above-mentioned technical problems, the object of the present invention is to provide a kind of artificial intelligence-based air tightness detection method and system for water turbidity detection, and the technical scheme adopted is as follows:

In a first aspect, an embodiment of the present invention provides an artificial intelligence-based air tightness detection method for detecting turbidity in water, the method comprising the following specific steps:

Obtain the image of the cylinder wall of the detection cylinder, detect the attached bubbles on the cylinder wall, and obtain the attached bubble map;

Based on the gray level, the dependent bubbles are clustered to obtain the bubble regions under each gray level; the sub-region with the largest density of dependent bubbles is selected in each bubble region;

Obtain the texture feature parameters and light-dark contrast feature parameters of each sub-region, wherein the texture feature parameters include energy, contrast, correlation and entropy, and the light-dark contrast feature parameters are obtained according to the difference between the sub-region grayscale and the cylinder wall grayscale;

Using AHP, select the sub-region that can best reflect water turbidity. Specifically, in the hierarchical structure model, the target layer is the sub-region that can best reflect water turbidity, the criterion layer is texture feature parameters and light-dark contrast feature parameters, and the scheme layer is the sub-region that can best reflect water turbidity. for each subregion;

The water turbidity is detected according to the texture feature parameters and light-dark contrast feature parameters of the selected sub-regions.

Further, Canopy clustering is combined with mean clustering to cluster attachment bubbles.

Further, in each bubble region, select the subregion with the largest density of attached bubbles, specifically:

Perform a traversal search based on the pixel coordinates of each attached bubble in the bubble area to obtain the largest convex hull composed of the attached bubbles in the bubble area; obtain the centroid of the convex hull, and calculate the distance from the centroid of the convex hull to the centroid of each attached bubble in the bubble area mean, take the centroid of any attached bubble in the bubble area as the initial circle center, and take the mean distance as the radius to obtain the initial circle;

The principal component direction is obtained according to several vectors whose initial circle center points to each attached bubble in the initial circle. The attached bubbles on both sides of the line passing through the initial circle center and perpendicular to the line where the principal component direction is located are the side with the larger number of attached bubbles as the bubbles to be traversed; according to The initial circle center points to the projection length of several vectors of the bubbles to be traversed in the direction of the principal component to determine the new circle center, and still takes the average distance as the radius to generate a new circle;

A new circle is continuously generated until it converges to the place where the density of the attached bubbles is the largest, and the position of the circle when it converges is the sub-region.

Further, the weight of each feature parameter can be obtained by using AHP, and based on each feature parameter and corresponding weight of the selected sub-region, the fuzzy comprehensive evaluation method is used to detect water turbidity.

Further, the acquisition of the cylinder wall image is specifically as follows: collecting a front view image of the detection cylinder, intercepting the cylinder wall region from the front view image, and obtaining the cylinder wall image through perspective transformation.

Further, a semantic segmentation network is used to detect the attached bubbles on the cylinder wall.

In a second aspect, another embodiment of the present invention provides an artificial intelligence-based air-tightness detection system for water turbidity detection. The system specifically includes a memory, a processor, and a system stored in the memory and running on the processor. A computer program, when the computer program is executed by a processor, realizes the steps of an artificial intelligence-based airtightness detection method for water turbidity detection.

The embodiment of the present invention has at least the following beneficial effects: the present invention can more sensitively and accurately obtain the observation factor reflecting the turbidity of the water body; each texture feature parameter obtained from the grayscale co-occurrence matrix is combined with the light-dark contrast feature parameter combined with the corresponding algorithm of fuzzy analysis to obtain the water body The process of turbidity, comprehensively analyzes the main change factors in the process of water body becoming turbid, and determines the weight value of each factor in the reaction of water body changes, and then obtains more accurate and effective detection results of water turbidity, and the present invention calculates the amount of Small.

Description of drawings

In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or in the prior art, the following briefly introduces the accompanying drawings that need to be used in the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description The drawings are only some embodiments of the present invention, and for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without any creative effort.

FIG. 1 is a schematic diagram of a cylinder wall area in an embodiment of the present invention.

Detailed ways

In order to further illustrate the technical means and effects adopted by the present invention to achieve the predetermined purpose of the invention, the following describes a method and system for detecting turbidity in water in air tightness detection based on artificial intelligence according to the present invention in conjunction with the preferred embodiments. , its specific implementation, structure, features and effects are described in detail as follows. In the following description, different "one embodiment" or "another embodiment" are not necessarily referring to the same embodiment. Furthermore, the particular features, structures, or characteristics in one or more embodiments may be combined in any suitable form.

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

The embodiments of the present invention take the following application scenarios as examples to illustrate the present invention:

The application scenario is: based on the visual perception method to detect the turbidity of the water body in the air tightness detection tank, so as to help managers to control the state of the water body in real time, and to replace the water body with high turbidity in time. When the water body is stable and the camera has a fixed angle of view, image acquisition is performed regularly, and the water turbidity detection result is obtained based on the acquired image. Therefore, the present invention needs to set the image acquisition interval T. For example, when T=3, it means that image acquisition is performed every 3 days. , the specific T value can be flexibly set according to the different environments of each air tightness device; it should be noted that the detection cylinder in the present invention needs to be a transparent detection cylinder, or several sides of which are transparent detection cylinders, namely The device to be inspected can be seen through the side of the inspection cylinder.

An embodiment of the present invention provides a method for detecting turbidity of water in air-tightness detection based on artificial intelligence, the method comprising the following steps:

In step S1, the image of the cylinder wall of the detection cylinder is acquired, and the detection of the attached bubbles on the cylinder wall is performed to obtain the attached bubble map.

(a) Collect the front view image of the detection cylinder. In the embodiment, the front view image is shown in FIG. 1 . The cylinder wall area is intercepted from the front view image, and the cylinder wall image is obtained through perspective transformation. Among them, in the embodiment, the side of the cylinder wall with the clearest and densest bubbles is selected, and the area on the cylinder wall above the device to be detected and below the water surface is the cylinder wall area, and in the embodiment, the cylinder wall area is the black rectangular frame on the right side in Figure 1. Area.

的矩阵，具体公式为：The purpose of perspective transformation is to reduce the bubble area error caused by different viewing angles. ,y') space mapping is based on the transformation of the four fixed vertices of the image, that is, the four corners of the cylinder wall area; the perspective transformation is achieved by matrix multiplication, using a

matrix, the specific formula is:

and then:

at last:

Among them, the first two rows of the matrix implement linear transformation and translation, and the third row is used to implement perspective transformation. So far, the obtained cylinder wall image after perspective transformation, the cylinder wall image is a front view image.

(b) Preferably, the embodiment utilizes a semantic segmentation network to detect the adhering air bubbles on the cylinder wall.

The semantic segmentation network is the structure of Encoder-Decoder, and the specific training content is as follows:

(1) Collect the images of facing the cylinder wall with attached bubbles to construct a training data set, and mark the training data set. The attached bubbles are marked as 1, and the others are marked as 0. Among them, 80% of the data is randomly selected as the training set, and the remaining 20% is used as the validation set.

(2) Input the training image and label data into the semantic segmentation network, the Encoder extracts the image features, and converts the number of channels into the number of categories to obtain the feature map; then the height and width of the feature map are transformed into the size of the input image through the Decoder , which outputs the class of each pixel.

(3) The Loss function is trained using the cross-entropy loss function.

Multiply the attached bubble segmentation map obtained by the semantic segmentation network with the cylinder wall image to obtain the attached bubble map including only the attached bubbles.

Step S2, clustering the dependent bubbles based on the gray level to obtain the bubble regions under each gray level category; in each bubble region, select the sub-region with the largest density of the dependent bubbles.

(a) Convert the attached bubble map to a grayscale image. Due to illumination, the grayscale of each bubble in the imaging is inconsistent. Therefore, in the present invention, a method combining Canopy and mean clustering is used to classify the grayscale of the bubbles ; The reason for using the Canopy clustering algorithm is that the algorithm does not need to specify the k value (that is, the number of clustering) in advance, so the data can be roughly clustered according to the algorithm, and then the mean clustering can be used to fine cluster after the k value is obtained. After the clustering is completed, the bubble area under each grayscale category is obtained.

(b) In each bubble region, select the subregion with the largest density of attached bubbles:

For each bubble area, perform a traversal search based on the pixel coordinates of each attached bubble in the bubble area to obtain the largest convex hull composed of the attached bubbles in the bubble area; obtain the centroid of the convex hull, and calculate the centroid of the convex hull into the bubble area The mean value of the distance between the centroids of each attached bubble, taking the centroid of any attached bubble in the bubble area as the initial circle center, and taking the mean distance as the radius, the initial circle is obtained.

The principal component direction is obtained according to several vectors whose initial circle center points to each attached bubble in the initial circle. The attached bubbles on both sides of the line passing through the initial circle center and perpendicular to the line where the principal component direction is located are the side with the larger number of attached bubbles as the bubbles to be traversed; according to The initial circle center points to the projected length of several vectors of the bubbles to be traversed in the direction of the principal component to determine the new circle center, and the average distance is still used as the radius to generate a new circle, that is, all the generated circles have the same size. Among them, as the vertical line of the line where the direction of the principal component is located, the projected lengths corresponding to the bubbles to be traversed on both sides of the vertical line are divided into positive and negative, that is, the projected lengths have two directions; the initial circle center points to several vectors to be traversed. The average projection length of the principal component direction can be obtained to obtain the average projection length, and at the same time, the direction of the average projection length can be obtained, and then the position of the endpoint of the average projection length can be obtained.

Continue to generate new circles until it converges to the place with the largest density of attached bubbles, and the position of the circle when it converges is the sub-region; after multiple iterations, the position of the new center of the circle does not change, then the circle region obtained based on the center of the circle is The sub-region with the highest density of attached bubbles.

So far, the sub-region with the largest density of attached bubbles in each bubble region can be obtained, and the sub-region is circular.

Step S3, acquiring the texture feature parameters and light-dark contrast feature parameters of each sub-region, wherein the texture feature parameters include energy, contrast, correlation and entropy, and the light-dark contrast feature parameters are obtained according to the difference between the sub-region grayscale and the cylinder wall grayscale. ;Using the analytic hierarchy process, select the sub-region that can best reflect water turbidity. Specifically, in the hierarchical structure model, the target layer is the sub-region that can best reflect water turbidity, and the criterion layer is texture feature parameters and light-dark contrast feature parameters. Layers are sub-regions.

(a) For each sub-region, calculate the corresponding grayscale co-occurrence matrix, and then obtain some eigenvalues of the matrix by calculating the gray-level co-occurrence matrix to represent the texture features of the image respectively; the reason for using the gray-level co-occurrence matrix is that it can It can comprehensively reflect the comprehensive information about the direction, adjacent interval, and change range of the image, so the subsequent fuzzy analysis method can more accurately reflect the change of the water body.

The embodiment selects the four most commonly used feature parameters of the gray level co-occurrence matrix to extract the texture features of the image. The four features and their representative meanings are as follows:

Energy γ; Energy is the sum of squares of each element of the gray co-occurrence matrix, also known as the second-order moment of angle.

Contrast ε; Contrast is the moment of inertia near the main diagonal of the grayscale co-occurrence matrix, which reflects how the values of the matrix are distributed, and reflects the clarity of the image and the depth of the texture grooves.

Correlation degree ϵ; the correlation degree reflects the similarity of the elements of the spatial grayscale co-occurrence matrix in the row or column direction, and reflects the correlation of the local grayscale of the image.

Entropy σ; Entropy reflects the randomness of the image texture. If all the values in the gray-level co-occurrence matrix are equal, it will get the maximum value; if the values in the co-occurrence matrix are not uniform, its value will become very small.

In addition to the above four characteristic parameters obtained from the grayscale co-occurrence matrix, the light-dark contrast characteristic parameter μ is also used in the present invention to represent the light-dark contrast between the bubbles in the density circle area and the cylinder wall background on the grayscale, because the more obvious the light-dark contrast is , it is easier to observe the changes caused by the turbidity of the water body. However, in the process of the water body becoming turbid gradually, the gray level of the bubbles and the background gray level will gradually become one area, and it is difficult to distinguish them in the end. The absolute value of the difference between the gray mean value and the gray mean value of the cylinder wall is obtained.

(b) The embodiment utilizes the Analytic Hierarchy Process, selects the sub-region that can best reflect the water turbidity, and carries out weight analysis to five characteristic parameters, specifically:

Establish a hierarchical structure model: the target layer is the sub-region that can best reflect the turbidity of water, and the criterion layer is the influencing factor, that is, texture feature parameters and light-dark contrast feature parameters, that is, energy, contrast, correlation, entropy, and light-dark contrast feature parameters. Layers are sub-regions.

Constructing a comparative judgment matrix: when determining the weights between each factor at each level, it is analyzed and determined through the comparison between two factors.

Single-level ranking and its consistency test: determine whether the ranking weight of factors at the same level for the relative importance of a factor in the previous level is correct and reasonable.

Hierarchical total ranking and its consistency check: Calculate whether the relative importance of all factors of a certain layer to the highest layer is reasonable.

Since the weight of each factor in analyzing the changes in water turbidity is uncertain, and the AHP is developed to solve such problems, the AHP can be used in the present invention to obtain the regions described in the above process. The weight of the density circle in response to the change of water turbidity, and the weight coefficient of the corresponding five factors in response to the change of water turbidity is determined by the maximum weight of the corresponding sub-region.

So far, the sub-regions that can best reflect the turbidity of water and the corresponding weights of each feature parameter can be obtained.

Step S4, the water turbidity is detected according to the texture feature parameters and the light-dark contrast feature parameters of the selected sub-region.

The weight of each characteristic parameter can be obtained by using AHP, and based on each characteristic parameter and corresponding weight of the selected sub-region, the fuzzy comprehensive evaluation method is used to detect the water turbidity; There are four turbidity grades: moderate turbidity and severe turbidity. Based on the fuzzy comprehensive evaluation method, the membership degree of each turbidity grade can be obtained. According to the maximum value of membership degree, the turbidity grade of the water in the detection tank can be obtained. When the detection result is moderate When it is turbid or severely turbid, the relevant personnel are reminded to change the water to prevent errors in the air tightness test results caused by the turbidity of the water body in the subsequent air tightness test.

Based on the same inventive concept as the above method embodiments, an embodiment of the present invention provides an artificial intelligence-based airtightness detection system for water turbidity detection. The system includes a memory, a processor, and a system stored on the memory and available at A computer program running on the processor, when the computer program is executed by the processor, realizes the steps of an artificial intelligence-based airtightness detection method for water turbidity detection.

It should be noted that: the above-mentioned order of the embodiments of the present invention is only for description, and does not represent the advantages and disadvantages of the embodiments. And the foregoing describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that in the embodiments and still achieve desirable results. Additionally, the processes depicted in the figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

Each embodiment in this specification is described in a progressive manner, and the same and similar parts between the various embodiments may be referred to each other, and each embodiment focuses on the differences from other embodiments.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection of the present invention. within the range.

Claims (7)

1. A method for detecting water turbidity in air tightness detection based on artificial intelligence is characterized by comprising the following steps:

acquiring a cylinder wall image of a detection cylinder, and detecting attached bubbles on the cylinder wall to obtain an attached bubble image;

clustering the attached bubbles based on the gray level to obtain bubble areas under each gray level category; selecting a sub-area with the largest attached bubble density in each bubble area;

acquiring texture characteristic parameters and light and shade contrast characteristic parameters of each subregion, wherein the texture characteristic parameters comprise energy, contrast, correlation and entropy, and the light and shade contrast characteristic parameters are obtained according to the difference value of the grayscale of the subregions and the grayscale of the cylinder wall;

selecting a subregion most capable of reacting to the turbidity of water by utilizing an analytic hierarchy process, specifically, selecting a target layer in a hierarchical structure model as the subregion most capable of reacting to the turbidity of water, a criterion layer as a texture characteristic parameter and a light and shade contrast characteristic parameter, and a scheme layer as each subregion;

and detecting the water turbidity according to the texture characteristic parameters and the light and shade contrast characteristic parameters of the selected sub-regions.

2. The method of claim 1, wherein Canopy clustering is combined with mean clustering to cluster dependent bubbles.

3. The method according to claim 2, characterized in that the sub-area with the highest attached bubble density is selected in each bubble area, specifically:

traversing search is carried out based on the coordinates of pixel points attached to the bubbles in the bubble area, and a maximum convex hull formed by the attached bubbles in the bubble area is obtained; acquiring the mass center of the convex hull, calculating the average distance from the mass center of the convex hull to the mass center of each attached bubble in the bubble area, taking the mass center of any attached bubble in the bubble area as an initial circle center, and taking the average distance as a radius to obtain an initial circle;

acquiring a principal component direction according to a plurality of vectors of the attached bubbles pointing to the initial circle from the initial circle center, wherein the attached bubbles on one side with a larger number of the attached bubbles are taken as bubbles to be traversed on two sides of a straight line which passes through the initial circle center and is vertical to the straight line of the principal component direction; determining a new circle center according to the projection length of a plurality of vectors of the initial circle center pointing to each bubble to be traversed in the principal component direction, and generating a new circle by still taking the distance mean value as the radius;

and continuously generating new circles until the new circles converge to the place where the density of the attached bubbles is maximum, wherein the positions of the circles during convergence are the sub-areas.

4. The method of claim 3, wherein the weights of the characteristic parameters are obtained by an analytic hierarchy process, and the detection of the turbidity of the water is performed by a fuzzy comprehensive evaluation process based on the characteristic parameters and the corresponding weights of the selected sub-regions.

5. The method as claimed in claim 4, wherein the cylinder wall image is obtained by acquiring an elevation image of the detection cylinder, cutting a cylinder wall area in the elevation image, and obtaining a cylinder wall image through perspective transformation.

6. The method of claim 1, wherein detecting attached bubbles on the cylinder wall is performed using a semantic segmentation network.

7. An artificial intelligence based water turbidity detection system in air tightness detection, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method according to any one of claims 1-6.

Images