Image perception based method for grading the level of deposition dust contamination

Abstract

The present application relates to a kind of image perception-based deposition dust pollution degree grading method, belong to image processing technical field.The method includes: first, design the marker for determining the pollution level of target area deposition dust, and obtain the image set of different clean degree of marker, on this basis, the marker segmentation model is obtained by using semantic segmentation method training;Second, arrange the marker in target area, obtain real-time monitoring image of target area, obtain the image of marker by marker segmentation model, perspective transformation algorithm and down / up sampling, record the reference image of marker, and determine the low value channel of marker, the luminance value of reference low value channel of marker, structure channel, the standard deviation of marker structure channel;On this basis, for the image to be determined, calculate the brightening parameter of marker image, structure parameter and detail parameter, finally, weighted target area deposition dust pollution degree grading parameter is obtained, and the pollution degree of target area deposition dust is graded according to its value.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology and relates to a method for classifying the degree of sedimentary dust pollution based on image perception. Background Technology

[0002] For enterprises involved in the production and processing of combustible dust, monitoring and controlling the amount of dust deposits in localized areas is crucial to ensuring production safety. Combustible dusts, such as sawdust, grain dust, and chemical powders, can form flammable mixtures with air under certain conditions, which can potentially ignite upon contact with an ignition source such as a spark or static electricity. Therefore, assessing the level of pollution from these dusts is a key step in preventing fires and explosions. Enterprises must implement effective monitoring measures to regularly test the dust concentration and deposition status in the environment.

[0003] Currently, most factories assess the degree of dust contamination in localized areas through manual visual inspection. This method is not only time-consuming and inefficient, but also highly susceptible to subjective factors, making it difficult to guarantee the consistency and accuracy of the assessment. In contrast, image processing technology offers a more comprehensive and dynamic monitoring method. Image monitoring systems can be integrated with existing safety monitoring systems, utilizing deployed video surveillance facilities to determine the dust deposition situation in contaminated areas based on real-time captured images, greatly improving the enterprise's safety early warning and accident response capabilities.

[0004] In the factory environment of processing enterprises, the similarity in color between accumulated dust and the surface of the area to be monitored makes it difficult for image processing algorithms to distinguish between dust and background, thus reducing detection accuracy. Meanwhile, changes in lighting conditions further affect the accuracy of image-based methods, making dust detection even more challenging. These problems not only reduce the precision and resolution of image-based measurements but also increase the complexity of data processing, posing a challenge to accurately determining the degree of dust contamination in a region. Therefore, effectively capturing the image features of deposited dust and researching and developing methods that can quickly and efficiently detect the degree of dust contamination in polluted areas are of great significance for improving the practicality of image processing technology in determining the degree of dust contamination in working environments. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method for classifying the degree of deposited dust pollution based on image perception, which can determine the degree of deposited dust pollution in a target area in real time based on captured images, thereby improving the enterprise's safety early warning and accident response capabilities.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A method for classifying the degree of sedimentary dust pollution based on image perception, the method comprising: S1. Design markers to determine the level of dust pollution in the target area, and obtain image sets of markers with different cleanliness levels. Use semantic segmentation methods to train a marker segmentation model. S2. Place markers in the target area, and collect and store the target area rating benchmark image while ensuring that the markers are not obstructed or contaminated by dust. S3. Input the target region rating benchmark image into the trained marker segmentation model to obtain the marker region benchmark map; S4. Perform perspective transformation and downsampling / upsampling operations on the landmarks in the landmark area reference map to obtain the landmark reference image; S5. Calculate the low-value channel of the marker, the brightness value of the low-value channel of the marker, the structure channel of the marker, and the standard deviation of the marker reference image based on the marker reference image; S6. Obtain real-time monitoring images of the target area and input them into the trained marker segmentation model to obtain a marker area map; S7. Perform perspective transformation and downsampling / upsampling operations on the landmarks in the landmark area map to obtain the landmark image; S8. Calculate the marker image brightening parameters by combining the marker image and the parameters obtained in step S5. H Structural parameters S And detailed parameters; S9, Image brightening parameters based on landmarks H Structural parameters S The classification parameters for the degree of sedimentary dust pollution in the target area are obtained by weighted calculation of detailed parameters. C ,according to C The value is used to classify the degree of dust pollution in the target area.

[0007] The beneficial effects of this invention are as follows: This invention can determine the degree of dust pollution in a target area in real time based on captured images. This invention has high judgment efficiency, short time consumption, little influence from subjective factors, and high consistency of assessment, which greatly improves the enterprise's safety early warning and accident response capabilities.

[0008] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0009] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1This is a flowchart illustrating the method described in this invention. Detailed Implementation

[0010] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0011] This invention proposes an image-aware-based method for classifying the degree of sedimentary dust pollution. The method first designs markers to determine the level of sedimentary dust pollution in a target area and acquires image sets of markers with different cleanliness levels. Based on this, a marker segmentation model is trained using semantic segmentation. Secondly, markers are deployed in the target area, and real-time monitoring images of the target area are acquired. P Marker images are obtained through marker segmentation models, perspective transformation algorithms, and downsampling / upsampling. I ; Acquire reference images of markers ,Depend on Determine the low-value channel of the marker D , Marker reference low channel brightness value Structural Channel V Standard deviation of marker structure channel Based on this, for the image to be judged I Calculate the highlighting parameters for the marker image. H Structural parameters S and detailed parameters , Finally, weighted parameters were used to determine the classification of dust pollution levels in the target area. C ,according to C The value of is used to classify the degree of dust pollution in the target area.

[0012] like Figure 1 As shown, the specific content of the method of the present invention includes: 1. Acquire images of deposited dust markers I 1) Establish a marker extraction model using semantic segmentation algorithms. ① Design markers to determine the level of dust pollution in a target area. The markers should have low reflectivity, simple and repetitive structures, such as equally spaced parallel yellow and black diagonal stripes. ② Construct a marker dataset, which should include images of markers with different cleanliness levels; ③ Use image labeling software (such as LabelMe) to label the markers in the dataset, and put the label set and the dataset into a semantic segmentation model (such as U-net) to train a marker extraction model.

[0013] 2) Obtain preliminary images of markers using a marker extraction model. ① In the target area where cleanliness assessment is required, set up the designed markers, and use a camera to record images of the target area. Under the condition that the markers are not obstructed and are not contaminated by dust, collect and store the baseline images. ;Will The landmark region baseline image is obtained from the input landmark extraction model. Perspective transformation and downsampling / upsampling operations are performed on the landmarks in the baseline image to make the landmark image size m×n×C, where m, n, and C are the pre-defined length, width, and number of color channels of the landmark image, respectively. This current landmark image is recorded as the landmark baseline image. .

[0014] ② Set the image acquisition interval T, and store the acquired image at intervals of time T. ; ③ Use the established marker extraction model to process the image Real-time target segmentation is performed to extract landmark region maps. The landmarks are then assessed: if a landmark is occluded, the current image is discarded; if the landmark is intact, perspective transformation and downsampling / upsampling operations are performed to give it a uniform size of m×n×C, resulting in the landmark image. .

[0015] 2. Calculate the highlighting parameters, structural parameters, and detail parameters of the landmark image. 1) Calculate the marker brightening parameters H ① Determine the low-value channel of the marker Solve separately The brightness values ​​of the R, G, and B channels are obtained as follows:

[0016]

[0017]

[0018] in , , They are respectively Medium pixels ( , The red, green, and blue channel values ​​of )m , n Given the length and width values ​​of the marker image, respectively, the baseline low-value channel of the marker and its corresponding brightness values ​​are as follows:

[0019]

[0020] ② Calculate the marker brightening parameters H For logo images Calculate the low-value channel brightness value:

[0021] in for Pixels in the low-value channel ( , The value of ) will highlight the parameter. H Represented as:

[0022] in, A The maximum value that can be taken for the dark channel pixels.

[0023] 2) Calculate the structural parameter S of the marker. ① Determine the structural channel of the marker Solve separately The standard deviations of the pixel values ​​in the R, G, and B channels are obtained as follows:

[0024]

[0025]

[0026] The marker structure channels and their corresponding standard deviations are as follows:

[0027]

[0028] ② Calculate the structural parameters of the marker image S For logo images Calculate the standard deviation of its structural channels:

[0029] for Pixels in the structure channel ( , The value of ) for If the structure channel brightness value is used, then the structure parameters are:

[0030] 3) Calculate the detailed parameters of the marker. ,

[0031] As dust pollution worsens, the number of straight-line contours of markers decreases, while the number of irregular curved contours increases. To measure the changes in marker contours, a straight-line contour loss parameter is defined. The parameter represents the degree of reduction in straight profiles and the increase in curved profiles. This indicates the degree of increase in the curve profile.

[0032] ① Add parameters to calculate curve profile

[0033] A. Based on the landmark map low value channel : First, perform a Gaussian filter on it. The Gaussian filter is expressed as:

[0034] Then the low-value channel pixel value after filtering .

[0035] B. Then, the Sobel operator is used to calculate... The gradient of the gradient in the x-direction:

[0036] In the y-direction:

[0037] Then gradient magnitude for: gradient direction for: .

[0038] C. Perform non-maximum suppression on the gradient magnitude value. That is, for each pixel, check its two adjacent pixels in the gradient direction. If the current pixel is not a local maximum, set it to 0. .

[0039] D. Select two curve contours to extract thresholds, using the higher threshold... (usually taken) and low threshold (usually taken) The gradient magnitude value is compared with a threshold; if the gradient magnitude value is greater than a threshold, the gradient magnitude value is considered greater than a threshold. It is then considered a strong edge, if in and The edges between these edges are considered weak edges. All weak edges connected to strong edges, along with strong edges, are retained to obtain the edge map. .

[0040] For logo images low value channel Similarly, using steps A through D above, we can obtain its edge map. The curve profile then increases as shown in the diagram:

[0041] The parameters for the curve profile are increased as follows: ,in B The pixel value is used to determine the outline.

[0042] ② Calculate the straight profile loss parameters

[0043] Reference map for markers structural channel : A. First, perform a Gaussian filter on it. The Gaussian filter is expressed as:

[0044] Then the low-value channel pixel value after filtering

[0045] B. Then, the Sobel operator is used to calculate... The gradient of the gradient in the x-direction:

[0046] In the y-direction,

[0047] Then gradient magnitude for: gradient direction for: .

[0048] C. Secondly, non-maximum suppression is applied to the gradient magnitude value. That is, for each pixel, its two adjacent pixels in the gradient direction are checked. If the current pixel is not a local maximum, it is set to 0, resulting in... .

[0049] D. Select two straight line contours to extract thresholds, using the higher threshold... (usually taken) and low threshold (usually taken) The gradient magnitude value is compared with a threshold; if the gradient magnitude value is greater than a threshold, the gradient magnitude value is considered greater than a threshold. It is then considered a strong edge, if in and The edges between these edges are considered weak edges. All weak edges connected to strong edges, along with strong edges, are retained to obtain the edge map. .

[0050] For logo images structural channel Similarly, using steps A through D above, we can obtain its edge map. The straight line profile loss map is then represented as:

[0051] Then the straight profile loss parameter .

[0052] 3. Classify the degree of dust pollution in the target area. The classification parameters for the degree of dust pollution in the target area are expressed as follows:

[0053] in Let be the weighting coefficient, satisfying (usually taken) , , , Then, based on the grading parameters... C Pollution levels can be categorized as follows: High level of pollution:

[0054] Medium to high pollution levels:

[0055] Medium level of pollution:

[0056] Low to medium level of pollution:

[0057] Low level of pollution:

[0058] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for classifying the degree of sedimentary dust pollution based on image perception, characterized in that: The method includes the following steps: S1. Design markers to determine the level of dust pollution in the target area, and obtain image sets of markers with different cleanliness levels. Use semantic segmentation methods to train a marker segmentation model. S2. Place markers in the target area, and collect and store the target area rating benchmark image while ensuring that the markers are not obstructed or contaminated by dust. S3. Input the target region rating benchmark image into the trained marker segmentation model to obtain the marker region benchmark map; S4. Perform perspective transformation and downsampling / upsampling operations on the landmarks in the landmark area reference map to obtain the landmark reference image; S5. Calculate the low-value channel of the marker, the brightness value of the low-value channel of the marker, the structure channel of the marker, and the standard deviation of the marker reference image based on the marker reference image; S6. Obtain real-time monitoring images of the target area and input them into the trained marker segmentation model to obtain a marker area map; S7. Perform perspective transformation and downsampling / upsampling operations on the landmarks in the landmark area map to obtain the landmark image; S8. Calculate the marker image brightening parameters by combining the marker image and the parameters obtained in step S5. H Structural parameters S And detailed parameters; S9, Image brightening parameters based on landmarks H Structural parameters S The parameters for classifying the degree of sedimentary dust pollution in the target area are obtained by weighted calculation of detailed parameters. C ,according to C The values ​​are used to classify the degree of dust pollution in the target area; among them, structural parameters S The parameters are calculated based on the standard deviation of pixel values ​​in the R, G, and B channels and the standard deviation of the structure channel of the marker image; the detail parameters include the straight profile loss parameter and the curve profile increase parameter. The straight profile loss parameter represents the degree of reduction of the straight profile of the marker, and the curve profile increase parameter represents the degree of increase of the curve profile of the marker.

2. The method for classifying the degree of sedimentary dust pollution according to claim 1, characterized in that: The low-value channel of the marker D Calculated using the following formula: In the formula, , , These represent the reference images of the markers. R , G , B Brightness values ​​of the three channels, , , These represent the pixels in the reference image of the marker ( , )of R , G , B Channel value, m , n These represent the length and width values ​​of the logo image, respectively. The reference low-value channel brightness value of the marker Calculated using the following formula: 。 3. The method for classifying the degree of sedimentary dust pollution according to claim 2, characterized in that: Marker image brightening parameters H Calculated using the following formula: In the formula, Image of a marker The low channel brightness value, A This indicates the maximum value that the dark channel pixels can take. express Pixels in the low-value channel ( , The value of ).

4. The method for classifying the degree of sedimentary dust pollution according to claim 1, characterized in that: The marker structure channel V Calculated using the following formula: In the formula, , , These represent the reference images of the markers. R , G , B Standard deviation of pixel values ​​in the three channels; The standard deviation of the reference image of the marker Calculated using the following formula: In the formula, , , These represent the pixels in the reference image of the marker ( , )of R , G , B Channel value, m , n These represent the length and width values ​​of the logo image, respectively. , , These represent the reference images of the markers. R , G , B Brightness values ​​for the three channels.

5. The method for classifying the degree of sedimentary dust pollution according to claim 4, characterized in that: Marker structural parameters S Calculated using the following formula: In the formula, Image of a marker The standard deviation of the structural channel express Pixels in the structure channel ( , The value of ) express Structure channel brightness value.

6. The method for classifying the degree of sedimentary dust pollution according to claim 1, characterized in that: The curve profile is increased with parameters The calculation process includes: 1) For the reference image of the marker low value channel First, perform a Gaussian filter on it, then the low-value channel pixel values ​​are... ,in Indicates a Gaussian filter; 2) Calculation The gradient is obtained to determine the gradient magnitude. Non-maximum suppression is applied to the gradient magnitude values. Specifically, for each pixel, its two adjacent pixels along the gradient direction are checked; if the current pixel is not a local maximum, it is set to 0. ; 3) Select two curve profiles to extract thresholds, with the higher threshold being the lower threshold. and low threshold The gradient magnitude value is compared with a threshold; if the gradient magnitude value is greater than a threshold, then... It is then considered a strong edge, if in and The edges between these edges are considered weak edges. All weak edges connected to strong edges, along with strong edges, are retained to obtain the edge map. ; 4) For the image of the marker low value channel Using steps 1) to 3) above, the edge map is obtained. Then the curve contour increase graph is obtained. for: The parameters for the curve profile are increased as follows: ,in B The pixel values ​​representing the outline. m , n These represent the length and width values ​​of the logo image, respectively.

7. The method for classifying the degree of sedimentary dust pollution according to claim 1, characterized in that: The straight profile loss parameter The calculation process includes: 1) For the reference image of the marker structural channel Perform a Gaussian filter on it, and the low-value channel pixel values ​​after filtering , Indicates a Gaussian filter; 2) Calculation The gradient is obtained to determine the gradient magnitude. Non-maximum suppression is applied to the gradient magnitude values. Specifically, for each pixel, its two adjacent pixels along the gradient direction are checked; if the current pixel is not a local maximum, it is set to 0. ; 3) Select two straight line contours to extract the threshold, and set the higher threshold. and low threshold The gradient magnitude value is compared with a threshold; if the gradient magnitude value is greater than a threshold, then... It is then considered a strong edge, if in and The edges between these edges are considered weak edges. All weak edges connected to strong edges, along with strong edges, are retained to obtain the edge map. ; 4) For the image of the marker structural channel Similarly, using steps 1) to 3) above, the edge map is obtained. Then the straight profile loss map Represented as: Straight profile loss parameters , m , n These represent the length and width values ​​of the logo image, respectively.

8. The method for classifying the degree of sedimentary dust pollution according to claim 1, characterized in that: In step S9, the classification parameters for the degree of dust pollution in the target area are... C Calculated using the following formula: In the formula, For the weighting coefficients, satisfying ; Add parameters to the curve profile. These are the parameters for the straight line profile loss.

Images