A dam safety detection method and system based on image recognition

Through image recognition-based methods, the dam image data is screened and analyzed, and the abnormal areas of the dam are identified, solving the problem of difficulty in efficiently monitoring the dam safety in the existing technology, and effectively detecting and identifying the abnormal situation of the dam is achieved.

CN118628478BActive Publication Date: 2024-11-22SHANDONG YELLOW RIVER RIVER AFFAIRS BUREAU SHANDONG YELLOW RIVER INFORMATION CENT
View PDF 5 Cites 0 Cited by

Patent Information

Application Number
CN202411057707.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-02
Publication Date
2024-11-22
Estimated Expiration
2044-08-02

AI Technical Summary

Technical Problem

During flood seasons and non-flood seasons, the flood pressure and water level changes of river water to the dam pose a challenge to the dam safety. It is difficult for the existing technology to efficiently monitor the abnormal situation of the dam, especially in the case of changes in light conditions and shadow shading.

Method used

By using an image recognition method, by collecting image data in the embankment area, filtering images with clearness meet the requirements, dividing detection areas, coarse matching and proportional relationship analysis, determining the region of interest, color balance and comparison, and identifying the embankment and use a pre-trained classifier model to identify the exception type.

Benefits of technology

The dam monitoring efficiency is improved and abnormal situations of the dam can be effectively detected, such as landslides, slope breaks, cracks, seepage and collapse, ensuring the safe operation of the dam.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118628478B_ABST
    Figure CN118628478B_ABST
Patent Text Reader

Abstract

The present application discloses a dam safety detection method and system based on image recognition, which involves data processing technology, including: screening out images whose clarity meets the requirements in the collected image data; performing rough matching according to the detection area of ​​the dam divided in advance to determine the image to be detected; for any image to be detected, determining the size of the pixel area of ​​the image area that matches the detection areas of the dam, and determining the relevant image area as the area of ​​interest; for any area of ​​interest, determining the color deviation of at least one reference position in the image to be detected and in the detection area divided in advance; balancing any area of ​​interest according to the color deviation; comparing any area of ​​interest with the corresponding detection area to detect whether any area of ​​interest is abnormal. The method of the present application is used to determine whether there is an abnormality in the dam, which can greatly improve the monitoring efficiency of the dam and ensure the safe operation of the dam.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a dam safety detection method and system based on image recognition. Background Art

[0002] According to the water level of the river, river basins are divided into flood season and non-flood season.

[0003] During the flood season, facing huge flood pressure, the river water carries a huge amount of sediment, the flow surges, and the rapid change of water level poses a severe challenge to the safety of the dam. During the non-flood season, the water volume is relatively stable, but the state of the dam still needs to be closely monitored to prevent disasters. Summary of the invention

[0004] The embodiment of the present application provides a dam safety detection method and system based on image recognition, which is used to perform partial area comparison after screening based on the collected image data of the dam area to determine whether there is any abnormality in the dam.

[0005] The present application embodiment proposes a dam safety detection method based on image recognition, comprising:

[0006] Collect image data of the dam area, and select images whose clarity meets the requirements from the collected image data;

[0007] For the image with the required clarity, a rough match is performed according to the detection area of ​​the dam divided in advance to determine the image to be detected, wherein the detection area is divided based on the image of the dam without shadow or obstruction collected under good lighting conditions, and multiple reference positions are configured for any detection area;

[0008] For any image to be detected, determining the size of the pixel area of ​​the image region matching each detection region of the dam;

[0009] Determine a first proportional relationship between a pixel area of ​​a matched image region in each to-be-detected image and a pixel area of ​​a corresponding detection region, and determine the relevant image region as a region of interest when the first proportional relationship is greater than a preset threshold;

[0010] For any region of interest, match at least one adjacent reference position according to the first proportional relationship, and determine a color deviation of the at least one reference position in the image to be detected and in a pre-divided detection area;

[0011] According to the color deviation, balancing any region of interest;

[0012] According to any balanced region of interest, a comparison is performed with a corresponding pre-divided detection area to detect whether any region of interest has an abnormality.

[0013] Optionally, for the image with the required clarity, performing a rough match according to the pre-divided detection area of ​​the dam includes:

[0014] Detect at least three reference positions that are not in a straight line from an image with a definition that meets the requirements and a corresponding pre-divided detection area;

[0015] Determining a second proportional relationship between the connection patterns of the at least three reference positions detected;

[0016] scaling the detection area according to the second proportional relationship;

[0017] Within a specified pixel distance range from the center of the connection pattern of the at least three reference positions, a rough match is performed on the scaled detection area in an image with a definition that meets the requirements.

[0018] Optionally, determining a color deviation of the at least one reference position in the image to be detected and in a pre-divided detection area includes:

[0019] Determine a first color range within the at least one reference position range based on a pre-divided detection area and the image to be detected; and

[0020] A second color range outside the at least one reference position range is determined, with the deviation between the detection area and the first color range of the image to be detected being used as the color deviation, and the deviation of the second color range being used as a correction deviation range.

[0021] Optionally, balancing any region of interest according to the color deviation includes:

[0022] Based on the at least one reference position in the image to be detected, color balancing the entire image to be detected according to the color deviation; and

[0023] After color balancing, when the image to be detected contains an area where the local brightness or contrast is greater than a preset threshold, the image to be detected is corrected in a direction of reducing the brightness or contrast according to the correction deviation range.

[0024] Optionally, the method further includes detecting a shadow area in the image to be detected in the following manner:

[0025] For any region of interest, extracting a pixel interval included in the any region of interest;

[0026] Based on the included pixel intervals, the pixel interval containing the most pixels is used as the dam pixel interval to screen out possible shadow areas;

[0027] For each possible shadow area, if there is a continuous pixel area connected to it and not belonging to itself and the dam pixel interval, it is determined to be a shadow area.

[0028] Optionally, comparing any balanced region of interest with a corresponding pre-divided detection region includes:

[0029] Extracting the boundary of the dam in any region of interest and the pre-divided detection area;

[0030] Determine a plurality of alignment sub-region pairs according to the relationship between the maximum boundary among the extracted boundaries and other boundaries;

[0031] The similarity between any pair of aligned sub-regions is calculated successively and sequentially;

[0032] When the calculated similarity is less than a preset threshold, it is determined that an abnormality exists in the corresponding local area of ​​the dam.

[0033] Optionally, for any region of interest including the shadow area, comparing the balanced region of interest with a corresponding pre-divided detection area further comprises:

[0034] At the corresponding position of the shadow area in the pre-divided detection area, the pixel values ​​of the associated shadow pixel range with a preset transparency are superimposed, and then any region of interest and the boundary of the dam in the pre-divided detection area are extracted, and a comparison is performed to determine whether there is an abnormality in the local area of ​​the dam.

[0035] Optionally, the identified local area with abnormality also includes:

[0036] Extracting image features of a local area of ​​any region of interest where an abnormality exists;

[0037] The extracted image features are input into the pre-trained classifier model to identify the abnormal type.

[0038] An embodiment of the present application also proposes a dam safety detection system based on image recognition, including a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps of the dam safety detection method based on image recognition as described above are implemented.

[0039] The method of the embodiment of the present application is based on the collected image data of the dam area, and performs partial area comparison after screening to determine whether there is any abnormality in the dam, which can greatly improve the monitoring efficiency of the dam and ensure the safe operation of the dam.

[0040] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0042] Figure 1 The figure is a schematic diagram of the basic process of the dam safety detection method based on image recognition in this embodiment. DETAILED DESCRIPTION

[0043] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0044] The present application embodiment proposes a dam safety detection method based on image recognition, such as Figure 1 As shown, the following steps are included:

[0045] In step S101, image data of the dam area is collected, and images with definition that meets the requirements are screened out from the collected image data. In some examples, the definition of the collected image data of the dam area can be calculated, and the images are sorted according to the definition, so as to screen out the images that meet the requirements.

[0046] In step S102, for the image with the required clarity, a rough match is performed according to the detection area of ​​the dam that is pre-divided to determine the image to be detected, wherein the detection area is divided based on the image of the dam without shadow or obstruction collected under good lighting conditions, and multiple reference positions are configured for any detection area. In some examples, the detection area of ​​the dam that is pre-divided can be based on the image of the dam in a normal state, such as the dam front, the dam surface, etc., collected under good lighting conditions and divided into multiple detection areas. For example, for the dam surface area, the divided multiple detection areas can have partial overlap, so as to facilitate the subsequent identification and matching steps. After obtaining the divided detection area, for the newly collected image later, the divided detection area is used for rough matching to determine whether the image collected later contains or contains that part of the relevant detection area. In some examples, the reference position can be determined according to the actual situation of the dam, and generally a fixed object with stable color value or a small selected area can be selected.

[0047] In step S103, for any image to be detected, the size of the pixel area of ​​the image region matching each detection region of the dam is determined.

[0048] In step S104, a first proportional relationship between the pixel area of ​​the image area matched in each image to be detected and the pixel area of ​​the corresponding detection area is determined, and when the first proportional relationship is greater than a preset threshold, the relevant image area is determined to be the region of interest. In a specific example, by detecting the pixel area of ​​the image area matched with each detection area in any image to be detected, and by judging the size relationship of the pixel area, some image areas with matching detection areas but small pixel area are filtered to improve the accuracy of recognition judgment in subsequent recognition.

[0049] In step S105, for any region of interest, at least one adjacent reference position is matched according to the first proportional relationship, and the color deviation of the at least one reference position in the image to be detected and in the pre-divided detection area is determined. In some examples, since the image collected later may have color deviation, the embodiment of the present application further determines the color deviation between any region of interest and the corresponding reference position and the corresponding reference position in the pre-divided detection area. According to the aforementioned example, if the reference position is a color-stable fixture such as a dam, the shooting condition of the image collected later can be determined based on the fixture at the reference position, and color balance can be performed based on the color deviation.

[0050] In step S106, any one of the regions of interest is balanced according to the color deviation.

[0051] In step S107, any balanced region of interest is compared with a corresponding pre-divided detection area to detect whether any region of interest has an abnormality.

[0052] In a specific example, the anomalies that may be contained in the dam include landslides, collapses, cracks, water seepage (humidity), pipe gushing (bubbling springs), collapse, etc., which may be different from the pre-divided detection area in the image. After color balancing, the present application can detect whether there is an anomaly in any area of ​​interest in the later collected image by comparing or matching the pre-divided detection area with any area of ​​interest.

[0053] The method of the embodiment of the present application is based on the collected image data of the dam area, and performs partial area comparison after screening to determine whether there is any abnormality in the dam, which can greatly improve the monitoring efficiency of the dam and ensure the safe operation of the dam.

[0054] In some embodiments, for an image with a definition that meets the requirement, performing a rough match according to a pre-divided detection area of ​​the dam includes:

[0055] At least three reference positions that are not in a straight line are detected from an image with a definition that meets the requirements and a corresponding pre-divided detection area.

[0056] A second proportional relationship between the connection patterns of the detected at least three reference positions is determined. In some examples, for example, three reference positions are selected to be connected to form a triangle shape, thereby determining a proportional relationship between two triangles.

[0057] The detection area is scaled according to the second proportional relationship.

[0058] Within the specified pixel distance range from the center of the connection pattern of the at least three reference positions, the zoomed detection area is roughly matched in the image with the required clarity. The partial area matching the detection area in the later collected image can be determined by zooming and matching.

[0059] In some embodiments, determining the color deviation of the at least one reference position in the image to be detected and in the pre-divided detection area includes:

[0060] Determine a first color range within the at least one reference position range based on a pre-divided detection area and the image to be detected; and

[0061] Determine a second color range outside the at least one reference position range, using the deviation between the detection area and the first color range of the image to be detected as the color deviation, and the deviation of the second color range as the correction deviation range. Based on the aforementioned embodiment, the reference position in the present application is a fixed object with a specified color stability. Under ideal image acquisition conditions, the color deviation of the reference position corresponding to the dam image acquired later should theoretically be within a small stable range compared to the dam image acquired earlier. In actual applications, there may be differences in brightness, contrast, saturation, etc. due to different acquisition conditions and parameters. In the embodiment of the present application, color balancing is performed in the latter example by determining the first color range and the second color range of the corresponding positions in the two images.

[0062] In some embodiments, balancing any one of the regions of interest according to the color deviation comprises:

[0063] Based on the at least one reference position in the image to be detected, the image to be detected as a whole is color balanced according to the color deviation. In a specific example, color balancing mainly involves overall adjustment of brightness, contrast, etc. to reduce the color deviation of the reference position in the region of interest acquired later.

[0064] After color balancing, if the image to be detected contains an area where the local brightness or contrast is greater than a preset threshold, the image to be detected is corrected in a direction of reducing the brightness or contrast according to the correction deviation range. Further, in general, the tolerance of the correction deviation range is larger than the color deviation. If the brightness after adjustment is too large, the contrast or brightness is reduced by correcting the deviation range.

[0065] In some embodiments, the method further includes detecting the shadow area in the image to be detected in the following manner:

[0066] For any region of interest, extracting a pixel interval included in the any region of interest;

[0067] Based on the included pixel intervals, the pixel interval containing the most pixels is used as the dam pixel interval to screen out the possible shadow area. That is, except for the dam pixel interval, the rest is the possible shadow area, which may include multiple discontinuous sub-areas.

[0068] For each possible shadow area, if there is a continuous pixel area connected to it and does not belong to itself and the dam pixel interval, it is determined to be a shadow area. Further, for a possible shadow area, such as a shadow of a telephone pole, by detecting a continuous image area connected to it and does not belong to itself and the dam pixel interval, it can be determined whether the possible shadow area is caused by an object, thereby determining the shadow area.

[0069] In some embodiments, comparing any balanced region of interest with a corresponding pre-divided detection region comprises:

[0070] Extract any region of interest and the boundary of the dam in the pre-divided detection area. In some examples, any region of interest will most likely include the front area and the surface area of ​​the dam. In this case, in this example, any region of interest can be divided again by extracting the boundary.

[0071] According to the relationship between the largest boundary among the extracted boundaries and other boundaries, multiple comparison sub-region pairs are determined, that is, according to the correspondence between the boundaries in the two sub-images, similar to the correspondence with the reference position, the boundary pairs containing the same dam area are taken as comparison sub-region pairs.

[0072] The similarity between any pairs of compared sub-regions is calculated sequentially and successively. In some examples, the similarity calculation may be performed using MSE, cosine similarity, and the like.

[0073] When the calculated similarity is less than a preset threshold, it is determined that an abnormality exists in the local area corresponding to the dam. In a specific example, theoretically, the image similarity in the same boundary should be at a high value position, for example, the similarity is greater than 95%. If it is lower than this value, there may be local abnormal points in the dam.

[0074] In some embodiments, for any region of interest including a shadow area, comparing the balanced region of interest with a corresponding pre-divided detection area further comprises:

[0075] At the corresponding position of the shadow area in the pre-divided detection area, the pixel values ​​of the associated shadow pixel range with a preset transparency are superimposed, and then any region of interest and the boundary of the dam in the pre-divided detection area are extracted, and a comparison is performed to determine whether there is an abnormality in the local area of ​​the dam.

[0076] In the embodiment of the present application, for the case of a shadow area, by superimposing color values ​​in the gray to black color range, for example, and by setting the transparency to 40%-70%, for example, the shadow area can be roughly simulated in the pre-divided detection area, and by further comparison, it can be determined whether there is an abnormality in the local area of ​​the dam.

[0077] In some embodiments, the identified local area with abnormalities also includes:

[0078] Extracting image features of a local area of ​​any region of interest where an abnormality exists;

[0079] The extracted image features are input into a pre-trained classifier model to identify the type of anomaly. Based on the above embodiment, it can be determined whether there is an abnormality in the local area of ​​the dam. For the specific type of abnormality in the local area of ​​the dam, such as pits, slope failure, cracks, water seepage (humidity), pipe gushing (bubbling springs), collapse, etc., the embodiment of the present application further uses a multi-classifier, such as SVM, to make a judgment, so as to further realize the judgment of the specific abnormality type and improve the effect of abnormality recognition.

[0080] An embodiment of the present application also proposes a dam safety detection system based on image recognition, including a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps of the dam safety detection method based on image recognition as described above are implemented.

[0081] In addition, although exemplary embodiments have been described herein, the scope includes any and all embodiments based on the present disclosure with equivalent elements, modifications, omissions, combinations (e.g., various embodiments intersecting schemes), adaptations or changes. It is not limited to the examples described in this specification or during the implementation of this application, and its examples will be interpreted as non-exclusive.

[0082] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of them) can be used in combination with each other. For example, those of ordinary skill in the art may use other embodiments when reading the above description.

[0083] The above embodiments are merely exemplary embodiments of the present disclosure. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the essence and protection scope of the present disclosure, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present invention.

Claims

1. A dam safety detection method based on image recognition, characterized in that: include: Collect image data of the dam area, and select images whose clarity meets the requirements from the collected image data; For the image with the required clarity, a rough match is performed according to the detection area of ​​the dam divided in advance to determine the image to be detected, wherein the detection area is divided based on the image of the dam without shadow or obstruction collected under good lighting conditions, and multiple reference positions are configured for any detection area; For any image to be detected, determining the size of the pixel area of ​​the image region matching each detection region of the dam; Determine a first proportional relationship between a pixel area of ​​a matched image region in each to-be-detected image and a pixel area of ​​a corresponding detection region, and determine the relevant image region as a region of interest when the first proportional relationship is greater than a preset threshold; For any region of interest, match at least one adjacent reference position according to the first proportional relationship, and determine a color deviation of the at least one reference position in the image to be detected and in a pre-divided detection area; According to the color deviation, balancing any region of interest; According to any balanced region of interest, a comparison is performed with a corresponding pre-divided detection area to detect whether any region of interest has an abnormality.

2. The dam safety detection method based on image recognition according to claim 1, characterized in that: For the image with the required definition, a rough matching is performed according to the detection area of ​​the dam divided in advance, including: Detect at least three reference positions that are not in a straight line from an image with a definition that meets the requirements and a corresponding pre-divided detection area; Determining a second proportional relationship between the connection patterns of the at least three reference positions detected; scaling the detection area according to the second proportional relationship; Within a specified pixel distance range from the center of the connection pattern of the at least three reference positions, a rough match is performed on the scaled detection area in an image with a definition that meets the requirements.

3. The dam safety detection method based on image recognition according to claim 1, characterized in that: Determining the color deviation of the at least one reference position in the image to be detected and in the pre-divided detection area includes: Determine a first color range within the at least one reference position range based on a pre-divided detection area and the image to be detected; and A second color range outside the at least one reference position range is determined, and a deviation between the detection area and the first color range of the image to be detected is used as the color deviation, and a deviation of the second color range is used as a correction deviation range.

4. The dam safety detection method based on image recognition according to claim 3 is characterized in that: Balancing any one region of interest according to the color deviation comprises: Based on the at least one reference position in the image to be detected, color balancing the entire image to be detected according to the color deviation; and After color balancing, when the image to be detected contains an area where the local brightness or contrast is greater than a preset threshold, the image to be detected is corrected in a direction of reducing the brightness or contrast according to the correction deviation range.

5. The dam safety detection method based on image recognition according to claim 4, characterized in that: It also includes detecting the shadow area in the image to be detected in the following manner: For any region of interest, extracting a pixel interval included in the any region of interest; Based on the included pixel intervals, the pixel interval containing the most pixels is used as the dam pixel interval to screen out possible shadow areas; For each possible shadow area, if there is a continuous pixel area connected to it and not belonging to itself and the dam pixel interval, it is determined to be a shadow area.

6. The dam safety detection method based on image recognition according to claim 5, characterized in that: Comparing any of the balanced regions of interest with the corresponding pre-divided detection regions includes: Extracting the boundary of the dam in any region of interest and the pre-divided detection area; Determine a plurality of alignment sub-region pairs according to the relationship between the maximum boundary among the extracted boundaries and other boundaries; The similarity between any pair of aligned sub-regions is calculated successively and sequentially; When the calculated similarity is less than a preset threshold, it is determined that an abnormality exists in the corresponding local area of ​​the dam.

7. The dam safety detection method based on image recognition according to claim 6, characterized in that: For any region of interest including the shadow area, comparing the balanced region of interest with the corresponding pre-divided detection area further includes: At the corresponding position of the shadow area in the pre-divided detection area, the pixel values ​​of the associated shadow pixel range with a preset transparency are superimposed, and then any region of interest and the boundary of the dam in the pre-divided detection area are extracted, and a comparison is performed to determine whether there is an abnormality in the local area of ​​the dam.

8. The dam safety detection method based on image recognition according to claim 7 is characterized in that: The identified local areas with abnormalities also include: Extracting image features of a local area of ​​any region of interest where an abnormality exists; The extracted image features are input into the pre-trained classifier model to identify the abnormal type.

9. A dam safety detection system based on image recognition, characterized in that: The method comprises a processor and a memory, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the steps of the dam safety detection method based on image recognition as described in any one of claims 1 to 8 are implemented.

Citation Information

Patent Citations

  • Amphibious intelligent underwater camera holder system

    CN109058710A

  • Method, device and equipment for measuring area of video image region and storage medium

    CN111062984A

  • Method for generating hydraulic engineering dam simulation three-dimensional model

    CN114445581A

  • Urban river big data monitoring liquid level alarm system and method

    CN116202590A

  • Dam leakage area identification method and device based on unmanned aerial vehicle remote sensing infrared thermal image color distance

    CN118115900A