A method for monitoring debris flow disasters

Abstract

The application discloses a kind of debris flow disaster monitoring methods, it is related to disaster monitoring technical field, comprising the following steps: S1, obtain scene image, and the denoising operation is carried out to scene image, generates standard scene image;S2, according to the screening weight of each row in standard scene image, obtains to be monitored area;S3, according to each monitoring pixel point RGB value in to be monitored area, extracts debris flow disaster pre-occurrence point.The present application does not need large-scale deployment sensor, can be to mountainous area and other complex terrain and sensor difficult to lay out the debris flow high-incidence area, realize long-term and stable monitoring, break through the topographical limit of traditional sensor monitoring.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of disaster monitoring, in particular to a debris flow disaster monitoring method. BACKGROUND

[0002] Debris flow is a sudden and extremely destructive geological disaster, which can easily cause serious threats to the safety of life and property of surrounding residents, and to the transportation and water conservancy infrastructure, etc. Therefore, timely and accurate monitoring of debris flow disaster is one of the key links in disaster prevention and mitigation.

[0003] At present, the monitoring methods of debris flow disaster mainly include two types of sensor-based monitoring and image-based monitoring. Among them, the sensor-based monitoring method (such as using pressure sensors, vibration sensors and other devices) needs to deploy a large number of sensors in the debris flow prone area, which not only has high deployment and maintenance costs, but also has limited monitoring range. For complex terrain mountainous areas and other high-risk debris flow scenes, it is difficult to arrange sensors, and the sensors are easily damaged by debris flow impact, rain erosion and other harsh environmental factors, resulting in monitoring interruption and difficulty in realizing long-term and stable continuous monitoring.

[0004] With the development of computer vision and image technology, image-based debris flow monitoring methods gradually attract attention due to their non-contact monitoring, wide coverage and real-time acquisition of scene visual information. However, the existing image-based monitoring technology still has obvious shortcomings: for example, the color features such as image pixel RGB values are not fully mined and utilized, making it difficult to meet the actual demand for recognition accuracy and timeliness. SUMMARY

[0005] In order to solve the above problems, the present application provides a debris flow disaster monitoring method.

[0006] The technical scheme of the present application is as follows:

[0007] S1, acquiring a scene image and performing a denoising operation on the scene image to generate a standard scene image;

[0008] S2, obtaining a to-be-monitored area according to the screening weight of each row in the standard scene image;

[0009] S3, extracting a debris flow disaster pre-occurrence point according to the RGB values of each monitoring pixel point in the to-be-monitored area.

[0010] Further, S2 includes the following sub-steps:

[0011] S21, calculating the screening weight of each row according to the RGB values of the pixel points in each row in the standard scene image;

[0012] S22, screening a reference pixel point set of each row according to the screening weight of each row;

[0013] S23, obtaining a to-be-monitored region of the standard scene image according to the reference pixel point set of each row.

[0014] The beneficial effect of the further scheme is that in the present application, the screening weight of each row is calculated first, then the reference pixel point set of each row is screened based on the weight, and finally the to-be-monitored region is obtained by integration, which refines the acquisition logic of the to-be-monitored region and improves the accuracy of region screening, thereby defining a more accurate range for subsequent pre-occurrence point extraction.

[0015] Further, in S21, the expression of the screening weight is as follows:

[0016] ;

[0017] wherein, represents the maximum RGB value of each row, represents the minimum RGB value of each row in the standard scene image, represents an exponential operation.

[0018] The beneficial effect of the further scheme is that in the present application, the ratio of the maximum RGB value and the minimum RGB value of each row reflects the color difference degree of the row of pixels (before the occurrence of a debris flow, the mixture of soil and stone will cause characteristic changes in the color difference within the row); by combining the linear term and the exponential term, the influence of such difference on the screening weight is quantified.

[0019] Further, in S22, the screening weight of each row is multiplied by the maximum RGB value of each row, and the pixel points with an RGB value less than the multiplication result are screened as the reference pixel point set.

[0020] The beneficial effect of the further scheme is that in the present application, the screening weight of each row reflects the potential degree of the row associated with the debris flow, and the maximum RGB value of each row is the brightest / brighest color feature of the row; the threshold obtained by multiplying the two can screen out pixel points that not only meet the weight feature but also are not the most extreme (avoiding interference) in the row, as the reference pixel point set, so that the reference pixel point set is more in line with the color law of the debris flow precursor.

[0021] Further, S23 includes the following sub-steps:

[0022] S231, in the standard scene image, except for the first row, the first column, the last row and the last column, extracting the RGB values of each reference pixel point in each row of the reference pixel point set and the RGB values of the corresponding 8-neighborhood to form a reference matrix of the reference pixel point;

[0023] S232, extract singular values of each reference matrix corresponding covariance matrix;

[0024] S233, calculate the monitoring threshold of the standard scene image according to the singular values of all covariance matrices;

[0025] S234, extract the minimum RGB value of the standard scene image, if the ratio between the minimum RGB value and the RGB value of the reference pixel point is greater than the monitoring threshold, the pixel point belongs to the region to be monitored.

[0026] The beneficial effects of the above further scheme are: in the present application, the RGB values of the reference pixel and its 8-neighborhood (mudslide is a regional disaster, in addition to the change of a single pixel, the surrounding pixels will also appear linkage change due to soil and stone diffusion, etc., and 8-neighborhood can cover the local area characteristics) are extracted to form a reference matrix. Singular values can reflect the principal component characteristics of the matrix, the monitoring threshold is calculated based on all singular values, and it is judged whether the pixel belongs to the region to be monitored, which greatly improves the accuracy of the region to be monitored.

[0027] Further, in S233, the monitoring threshold of the standard scene image The expression is:

[0028] ;

[0029] Wherein, represents the th singular value, represents the number of singular values contained in all covariance matrices, represents the standard deviation of all singular values.

[0030] The beneficial effects of the above further scheme are: in the present application, singular values reflect the core characteristics of the reference matrix (pixel neighborhood), and the standard deviation can buffer the abnormal fluctuations of singular values. Combined with the core characteristics of singular values and the anti-interference of standard deviation, the calculation of the monitoring threshold is more robust.

[0031] Further, S3 includes the following sub-steps:

[0032] S31, in the region to be monitored, extract the median of all monitoring pixel point RGB values;

[0033] S32, calculate the updated color degree of the monitoring pixel point according to the median;

[0034] S33, take the pixel point with RGB value greater than the average value of all updated color degrees in the standard scene image as the mudslide disaster pre-occurrence point.

[0035] The beneficial effect of the further scheme is that in the present application, the median can reflect the intermediate level of the pixel color in the region, the color of the pre-occurrence point of the debris flow usually deviates from the intermediate level, and the pixel point with the RGB value greater than the updated color degree average is determined as the pre-occurrence point (before the occurrence of the debris flow, the color of the pre-occurrence point will become brighter / brighter due to soil and stone factors, so as to exceed the average), thereby avoiding the one-sidedness of the judgment by using a single RGB value.

[0036] Further, in S32, the updated color degree of the monitoring pixel point is monitored The expression is:

[0037] ;

[0038] Wherein, The RGB value of the monitoring pixel point is represented by R, G and B, The median is represented by M, A random number between 0 and 1 is represented by R, The maximum value is represented by max, The minimum value is represented by min.

[0039] The beneficial effect of the present application is that the present application reduces the monitoring range in multiple dimensions through row weight calculation, reference pixel point set screening, neighborhood matrix and singular value analysis, excludes irrelevant regional interference, makes the to-be-monitored region more consistent with the essential characteristics of the regional changes of the debris flow, accurately captures the key changes of the pixel color deviating from the regional normal state and the enhanced characteristics before the occurrence of the debris flow, and greatly reduces the misjudgment and omission. The present application does not need to deploy a large number of sensors, can realize long-term and stable monitoring in the debris flow high-incidence area with complex terrain such as mountains and other terrains where sensors are difficult to deploy, and breaks through the terrain limitation of traditional sensor monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 The flowchart of the debris flow disaster monitoring method. DETAILED DESCRIPTION

[0041] The embodiments of the present application will be further described below with reference to the accompanying drawings.

[0042] As shown in Figure 1 The present application provides a debris flow disaster monitoring method, comprising the following steps:

[0043] S1, acquiring a scene image and performing a denoising operation on the scene image to generate a standard scene image;

[0044] S2, obtaining a to-be-monitored region according to the screening weight of each row in the standard scene image;

[0045] S3, extracting a debris flow disaster pre-occurrence point according to the RGB value of each monitoring pixel point in the to-be-monitored region.

[0046] In the embodiment of the present application, S2 comprises the following sub-steps:

[0047] S21, calculating the screening weight of each row according to the pixel point RGB value of each row in the standard scene image;

[0048] S22, screening the reference pixel point set of each row according to the screening weight of each row;

[0049] S23, obtaining the to-be-monitored region of the standard scene image according to the reference pixel point set of each row.

[0050] In the present application, the screening weight of each row is calculated first, and then the reference pixel point set of each row is screened based on the weight, and finally the to-be-monitored region is integrated to refine the acquisition logic of the to-be-monitored region and improve the accuracy of region screening, thereby delineating a more accurate range for subsequent pre-occurrence point extraction.

[0051] In the embodiment of the present application, in S21, the expression of the screening weight is as follows:

[0052]

[0053] wherein, represents the maximum RGB value of each row, represents the minimum RGB value of each row in the standard scene image, represents the exponential operation.

[0054] In the present application, the ratio of the maximum RGB value and the minimum RGB value of each row reflects the color difference degree of the row of pixels (before the occurrence of a debris flow, the mixing of soil and stone will cause characteristic changes in the color difference within the row); by combining the linear term and the exponential term, the influence of such difference on the screening weight is quantified.

[0055] In the embodiment of the present application, in S22, the screening weight of each row is multiplied by the maximum RGB value of each row, and the pixel points with an RGB value less than the multiplication result are screened as the reference pixel point set.

[0056] In the present application, the screening weight of each row reflects the potential degree of the row associated with a debris flow, and the maximum RGB value of each row is the brightest / brighest color feature of the row; the threshold obtained by multiplying the two can screen out pixel points whose color intensity not only conforms to the weight feature but also is not the most extreme (to avoid interference) in the row, as the reference pixel point set, so that the reference pixel point set is more in line with the color law of the debris flow precursor.

[0057] In the embodiment of the present application, S23 comprises the following sub-steps:

[0058] ​S231, in the standard scene image, except the first row, the first column, the last row and the last column, the RGB value of each reference pixel point in each row of reference pixel point set and the RGB value of the corresponding 8-neighborhood are extracted to form a reference matrix of the reference pixel point;

[0059] S232, the singular value of the covariance matrix corresponding to each reference matrix is extracted;

[0060] S233, the monitoring threshold of the standard scene image is calculated according to the singular values of all covariance matrices;

[0061] S234, the minimum RGB value of the standard scene image is extracted, if the ratio between the minimum RGB value and the RGB value of the reference pixel point is greater than the monitoring threshold, the pixel point belongs to the region to be monitored.

[0062] In the application, the RGB values of the reference pixel and its 8-neighborhood (mudslide is a regional disaster, in addition to the change of a single pixel, the surrounding pixels will also change in linkage due to soil and stone diffusion, and the 8-neighborhood can cover the local regional characteristics) are extracted to form a reference matrix. The singular value can reflect the principal component characteristics of the matrix, the monitoring threshold is calculated based on all singular values, and whether the pixel belongs to the region to be monitored is judged, so that the accuracy of the region to be monitored is greatly improved.

[0063] In the embodiment of the application, in S233, the expression of the monitoring threshold of the standard scene image is:

[0064] ;

[0065] Wherein, the singular value of the first row is represented by the singular value of the first row, the singular value of the last row is represented by the singular value of the last row, the singular value of the last column is represented by the singular value of the last column, and the standard deviation of all singular values is represented by the standard deviation of all singular values.

[0066] In the application, the singular value reflects the core characteristics of the reference matrix (pixel neighborhood), and the standard deviation can buffer the abnormal fluctuation of the singular value, so that the calculation of the monitoring threshold is more robust in combination of the core characteristics of the singular value and the anti-interference of the standard deviation.

[0067] In the embodiment of the application, S3 includes the following sub-steps:

[0068] S31, in the region to be monitored, the median of all monitoring pixel point RGB values is extracted;

[0069] S32, the updated color degree of the monitoring pixel point is calculated according to the median;

[0070] S33, the pixel point with the RGB value greater than the average value of all updated color degrees in the standard scene image is taken as the mudslide disaster pre-occurrence point.​

[0071] In the present application, the median can reflect the intermediate level of the color of the pixels in the region, and the color of the pre-occurrence point of the debris flow usually deviates from the intermediate level. The pixel point with the RGB value greater than the average of the updated color degree is determined as the pre-occurrence point (before the occurrence of the debris flow, the color of the pre-occurrence point will become brighter / brighter due to soil and stone factors, so as to exceed the average), avoiding the one-sidedness of the judgment by using a single RGB value.

[0072] In the embodiment of the present application, in S32, the updated color degree of the monitoring pixel point is monitored The expression is:

[0073] ;

[0074] Among them, The RGB value of the monitoring pixel point is represented by R, G and B, The median is represented by M, A random number between 0 and 1 is represented by R, The maximum value is represented by max, The minimum value is represented by min.

[0075] Those skilled in the art will appreciate that the embodiments described herein are intended to help the reader understand the principles of the present application and should be understood as not limiting the scope of protection of the present application to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the present application without departing from the essence of the present application, and these modifications and combinations are still within the scope of protection of the present application.

Claims

1. A method for monitoring debris flow disasters, characterized in that, The method comprises the following steps: S1, acquiring a scene image and performing a denoising operation on the scene image to generate a standard scene image; S2, obtaining a to-be-monitored region according to a screening weight of each row in the standard scene image; S3, extracting a debris flow disaster pre-occurrence point according to an RGB value of each monitoring pixel point in the to-be-monitored region; The S2 comprises the following sub-steps: S21, calculating a screening weight of each row according to an RGB value of each pixel point in the standard scene image; S22, screening a reference pixel point set of each row according to the screening weight of each row; S23, obtaining a to-be-monitored region of the standard scene image according to the reference pixel point set of each row; In the S21, the screening weight The expression is: ； wherein, represents the maximum RGB value of each row, represents the minimum RGB value of each row in the standard scene image, represents an exponential operation; In the S22, the screening weight of each row is multiplied by a maximum RGB value of each row, and a pixel point with an RGB value smaller than a multiplication result is screened as the reference pixel point set; The S23 comprises the following sub-steps: S231, in the standard scene image, except for a first row, a first column, a last row and a last column, extracting an RGB value of each reference pixel point in each reference pixel point set and an RGB value of a corresponding 8-neighbor domain to form a reference matrix of the reference pixel point; S232, extracting a singular value of a covariance matrix corresponding to each reference matrix; S233, calculating a monitoring threshold of the standard scene image according to the singular value of all covariance matrices; S234, extracting a minimum RGB value of the standard scene image, and if a ratio between the minimum RGB value and an RGB value of the reference pixel point is greater than the monitoring threshold, the pixel point belongs to the to-be-monitored region; In the S233, the monitoring threshold value of the standard scene image The expression is: ； wherein, represents the number of singular values, represents the number of singular values contained by all covariance matrices, represents the standard deviation of all singular values;​ The S3 comprises the following sub-steps: S31, in the to-be-monitored region, extracting a median of RGB values of all monitoring pixel points; S32, calculating an updated color degree of the monitoring pixel point according to the median; S33, taking a pixel point with an RGB value greater than an average value of all updated color degrees in the standard scene image as the debris flow disaster pre-occurrence point; In the S32, the updated color degree of the pixel point is monitored The expression is: ； wherein, represents monitoring the pixel point RGB value, represents the median, represents a random number between 0-1, represents taking the maximum value, represents taking the minimum value.

Images