A slope deformation monitoring method based on image recognition

Abstract

The present invention relates to the field of image processing technology, and more specifically, the present invention relates to a slope deformation monitoring method based on image recognition, including: obtaining a slope interference phase map; segmenting the slope interference phase map to obtain multiple sub-regions, obtaining local differences in phase points in the sub-regions; determining the degree of difference in phase values ​​corresponding to the phase points and the phase differences in the sub-regions; performing regional segmentation; determining the distribution similarity and overall similarity of adjacent sub-regions when the segmentation is completed; calculating the comprehensive similarity of the sub-regions when the segmentation is completed; performing regional merging; filtering according to the size of the standard deviation of the final sub-region to achieve slope deformation monitoring. The present invention dynamically determines whether to perform regional segmentation and merging, and more accurately adapts to the spatial variability of deformation patterns in the slope interference phase map than the fixed threshold method, effectively improves the effect of quadtree filtering, and makes slope deformation monitoring more accurate.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a slope deformation monitoring method based on image recognition. Background Art

[0002] Ground-based synthetic aperture (SAR) radar has a lot of room for development in slope monitoring. SAR radar emits microwave pulses and receives echo signals, and uses Doppler frequency changes to extract ground object information. In slope monitoring, SAR radar can obtain deformation information on the slope surface, and extract tiny deformation information on the slope surface in combination with interferometric measurement technology to obtain interference phase images. In order to obtain higher image quality, it is often necessary to filter and optimize the interference phase image to improve the signal-to-noise ratio of the interference phase image and reduce the number of residual points in the interference phase image, which helps to better perform phase unwrapping, improve the efficiency and accuracy of slope monitoring, and provide strong support for the safety and stability of slope engineering. Quadtree filtering is a spatial filtering algorithm based on recursive segmentation, which is particularly suitable for images or phase images with large flat areas and complex edges. In essence, it recursively divides the image space into smaller areas, and performs different degrees of smoothing on areas of different levels, which is conducive to the processing of interference phase images, because phase images often contain changes of multiple scales. Quadtree filtering can be processed according to different local features to reduce unnecessary detail loss.

[0003] The patent document with announcement number CN111598788B discloses a single image defogging method based on quadtree decomposition and non-local prior, and the steps include: step 1, using an image acquisition device to obtain a haze image I(x) that needs to be defogged, and using a quadtree decomposition method to divide the haze image I(x) into different blocks; step 2, estimating the rough transmittance of the haze image based on non-local prior; step 3, performing a morphological closing operation on the rough transmittance obtained in step 2; step 4, using a smoothing filter to obtain a fine transmittance; step 5, restoring a haze-free image according to an atmospheric scattering model.

[0004] However, the above patent document does not solve the problem that in the process of quadtree filtering, the threshold setting of region division is the key factor in determining the filtering effect, and the deformation of the slope is often irregular and will be affected by various factors such as geological conditions and environmental changes. Therefore, the deformation pattern appearing in the slope interference phase map also has a certain spatial variability. The fixed region segmentation and merging threshold will affect the filtering effect of the final slope interference phase map, and thus affect the accuracy of the slope deformation monitoring results. Summary of the invention

[0005] In order to solve the above-mentioned problem that in the process of quadtree filtering, since the threshold setting of region division is the key factor determining the filtering effect, and the deformation of the slope is often irregular and will be affected by multiple factors, the deformation mode appearing in the slope interference phase map has a certain spatial variability, and the fixed region segmentation and merging threshold will affect the filtering effect of the final slope interference phase map, and further affect the accuracy of the deformation monitoring result of the slope, the present invention proposes a slope deformation monitoring method based on image recognition, which includes the following steps:

[0006] The collected slope interference phase image is divided into regions and filtered by using quadtree filtering, including: performing regional segmentation on the slope interference phase image to obtain multiple sub-regions, obtaining local differences of phase points in the sub-regions, and the local differences are used to determine the data deviation of the phase points in the corresponding data set; determining the difference expression degree of phase values ​​corresponding to the phase points according to the local differences of the phase points in the sub-regions; taking the difference between the phase values ​​corresponding to the maximum and minimum difference expression degrees as the phase difference of the sub-regions; in response to the phase difference being greater than a preset segmentation threshold, performing the next regional segmentation on the sub-region, otherwise, stopping the regional segmentation, and obtaining multiple sub-region; for the target sub-region and the adjacent sub-region in the multiple sub-regions at the end of segmentation, curve fitting is performed on the two sub-regions, and the distribution similarity of the two sub-regions is determined according to the area of ​​the figure formed by the fitting curves of the two sub-regions; according to the distribution similarity and the phase difference of the two sub-regions, the overall similarity of the two sub-regions is determined; the overall similarity of the target sub-region and all adjacent sub-regions is weighted averaged to obtain the comprehensive similarity of the target sub-region; according to the size of the comprehensive similarity, the regions are merged to obtain multiple final sub-regions; according to the size of the standard deviation of the final sub-region, filtering is performed to realize the monitoring of slope deformation.

[0007] Through multiple regional segmentation and merging of quadtree filtering, it can automatically adapt to the deformation characteristics of the slope, dynamically adjust the segmentation threshold according to the phase difference of different regions, avoid excessive or insufficient regional segmentation that may be caused by fixed thresholds, and thus ensure the accuracy of the filtering effect; since the deformation of the slope is usually non-uniform and irregular, by analyzing the local differences and phase changes of each sub-region, it can accurately capture the changes in the deformation area, and dynamically adjust the regional segmentation and merging according to these changes to ensure accurate monitoring of slope deformation; when performing curve fitting and similarity evaluation on the target sub-region and the adjacent region, the region is merged by calculating the similarity weighted average, which further improves the accuracy of filtering, can effectively reduce over-refined regional segmentation, avoid the generation of too many meaningless small regions, and ensure the stability of the final result; the final regional merging and filtering results are based on the similarity and standard deviation calculation of the region, which can effectively remove noise and improve the detection accuracy of slope deformation; by dynamically adjusting the standard deviation threshold, different types of deformation modes can be accurately monitored, avoiding the errors caused by fixed thresholds, and further improving the accuracy of monitoring results.

[0008] Furthermore, the local difference satisfies the following relationship:

[0009] ; In the formula, For the In the sub-area The local difference of the phase points, For the In the sub-area The phase value of the phase point, For the In the sub-area The phase point belongs to the The phase value of the phase point, For the The number of phase points in the data set to which the phase point belongs, is the absolute value symbol.

[0010] By calculating the difference between a phase point and other phase points in the data set to which it belongs, the details of local phase changes can be effectively captured, thereby improving the sensitivity to slope deformation; by weighted summation of the differences of other phase points around the phase point, the phase changes in the local area can be comprehensively considered, thereby reducing the impact of single phase point outliers on the filtering effect; by processing the absolute value of the phase point difference and normalizing it based on the number of other phase points around the phase point, a more stable and adaptive measurement method is provided.

[0011] Furthermore, the difference expression degree satisfies the following relationship:

[0012] ; In the formula, For the Phase value in sub-region The degree of difference in performance, For the In the sub-area The phase value is The local difference of the phase point, For the The phase value in the sub-region is The number of phase points.

[0013] By calculating the average value of local differences and comprehensively considering the differences between different phase points in the sub-area, the stability of the overall monitoring results is improved and the interference of outliers is reduced; by independently evaluating different phase values, it is helpful to identify different deformation modes in the area and conduct more accurate analysis; the degree of difference expression can help identify and analyze deformation modes, thereby providing support for subsequent processing and decision-making.

[0014] Furthermore, the distribution similarity satisfies the following relationship:

[0015] ; In the formula, The first sub-regions and The distribution similarity of adjacent sub-regions, The first sub-regions and The area of ​​the figure formed by the fitting curves of adjacent sub-regions, The first The fitting curve of the sub-region is The first The sub-region The fitting curves of adjacent sub-regions are and The first sub-regions and The maximum and minimum phase values ​​in adjacent sub-regions, is the natural exponential function, is the definite integral symbol, is the integration variable The differential of is the absolute value symbol.

[0016] By calculating the distribution similarity, the differences between different segmented regions can be accurately measured; by considering the maximum and minimum ranges of the phase values, the characteristics of different regions can be adapted to avoid deviations caused by regional differences; by integrating the differences between the fitting curves, the distribution characteristics of adjacent regions can be more accurately compared, thereby improving the accuracy of the analysis; distribution similarity helps to evaluate and optimize the quality of regional segmentation, ensure that the segmented regions have reasonable similarity, and improve the reliability of subsequent analysis.

[0017] Furthermore, the overall similarity satisfies the following relationship:

[0018] ; In the formula, The first sub-regions and The overall similarity of adjacent sub-regions, The first sub-regions and The distribution similarity of adjacent sub-regions, The first The phase difference corresponding to each sub-region is The first The sub-region The phase difference corresponding to the adjacent sub-regions.

[0019] By comprehensively considering distribution similarity and phase difference, a more comprehensive regional similarity measure is provided for overall similarity; when calculating overall similarity, the distribution characteristics and phase difference of the region are balanced to ensure the reasonable integration of the two and improve the accuracy of analysis; the introduction of phase difference effectively corrects the differences between regions, avoids the influence of extreme differences on the similarity measure, and makes the result more stable; overall similarity can more sensitively reflect the actual differences between regions when the phase difference is significant, which helps to compare adjacent regions more accurately.

[0020] Furthermore, the region merging includes: in response to the comprehensive similarity being greater than the average of the comprehensive similarities of all sub-regions at the end of segmentation, merging the target sub-region with an adjacent sub-region with the highest overall similarity; otherwise, stopping the region merging.

[0021] Furthermore, the filtering includes: in response to a phase value of a phase point being greater than a standard deviation of a phase value of the corresponding final sub-region, identifying the phase point as a pixel gross error point and removing it, thereby obtaining a filtered slope interference phase map.

[0022] Furthermore, the monitoring of slope deformation includes: using phase unwrapping to obtain the deformation phase of the filtered slope interference phase image, and then converting the deformation phase into a deformation image and outputting it, thereby completing the slope deformation monitoring based on image recognition.

[0023] The present invention has the following beneficial effects:

[0024] By obtaining the local difference of phase points in the sub-region, the phase difference is determined, and then the decision on whether to perform regional segmentation is dynamically made. Compared with the fixed threshold method, it can more accurately adapt to the spatial variability of the deformation mode in the slope interference phase map, and effectively improve the effect of quadtree filtering; the distribution similarity is determined based on the area of ​​the figure composed of the fitting curves of the two regions, and the overall similarity is obtained by combining the phase difference, and then the comprehensive similarity is obtained by weighted average, and the regions are merged according to the comprehensive similarity. Finally, filtering is performed according to the final sub-region standard deviation, so that the processing of the slope interference phase map is more in line with the actual situation, and the influence of poor filtering on the accuracy of the slope deformation monitoring results is greatly reduced; this method fully considers the characteristics of slope deformation being affected by multiple factors and irregular, does not rely on fixed regional segmentation and merging thresholds, can be dynamically adjusted according to the characteristics of the image itself, and has stronger adaptability to complex and changeable slope deformation conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0026] Figure 1 The present invention is a flowchart of a slope deformation monitoring method based on image recognition according to an embodiment of the present invention.

[0027] Figure 2 It is a schematic diagram of region segmentation of a slope deformation monitoring method based on image recognition according to an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0029] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0030] See also Figure 1 , which shows a flow chart of a slope deformation monitoring method based on image recognition provided by an embodiment of the present invention, the method comprising the following steps:

[0031] S1: Obtain slope interferometry phase map.

[0032] It should be noted that in the process of quadtree filtering, the threshold setting of regional segmentation and merging is the key factor in determining the filtering effect. The deformation of the slope is often irregular and will be affected by many factors such as geological conditions and environmental changes. The deformation pattern that may appear in its interference phase map also has a certain spatial variability. Fixed regional merging and segmentation thresholds will affect the final filtering effect, and then affect the accuracy of the slope deformation monitoring results. Adjusting the regional segmentation and merging strategy according to the application scenario and the characteristics of the region itself can solve the above problems well.

[0033] The SAR radar technology is used to collect the interference phase image of the slope, and then the quadtree algorithm is used to divide the slope interference phase image into regions and filter it. The specific steps are as follows.

[0034] S2: Segment the slope interference phase image to obtain multiple sub-regions, and obtain the local differences of the phase points in the sub-regions.

[0035] It should be noted that the principle of the interference phase image is to map the phase value to a grayscale image drawn in the range of 0 to 255. The essence of each pixel is actually phase information, so phase points are used to replace pixels in the subsequent analysis process, but the analysis process still belongs to the field of image processing.

[0036] The slope interference phase image is segmented into multiple sub-regions, and local differences of phase points in the sub-regions are obtained. The local differences are used to determine the data deviation of the phase points in the corresponding data set.

[0037] Implementers can set the range of the dataset according to the specific implementation situation, for example, 8 neighborhoods.

[0038] Specifically, the local difference satisfies the following relationship:

[0039] ;

[0040] In the formula, For the In the sub-area The local difference of the phase points, For the In the sub-area The phase value of the phase point, For the In the sub-area The phase point belongs to the The phase value of the phase point, For the The number of phase points in the data set to which the phase point belongs, is the absolute value symbol.

[0041] in, Indicates In the sub-area The mean difference between a phase point and the phase points in the data set to which it belongs, that is, the data deviation. The larger the value, the In the sub-area The higher the local difference of each phase point.

[0042] S3: Determine the difference in the phase value corresponding to the phase point and the phase difference of the sub-region.

[0043] According to the local differences of the phase points in the sub-regions, the degree of difference in the phase values ​​corresponding to the phase points is determined.

[0044] Specifically, the difference expression degree satisfies the following relationship:

[0045] ;

[0046] In the formula, For the Phase value in sub-region The degree of difference in performance, For the In the sub-area The phase value is The local difference of the phase point, For the The phase value in the sub-region is The number of phase points.

[0047] in, Indicates The phase value in the sub-region is The local difference mean of the phase point. The larger the value, the smaller the phase value. The higher the degree of difference.

[0048] Specifically, the phase difference of the sub-regions satisfies the following relationship:

[0049] ;

[0050] In the formula, For the The phase value corresponding to the maximum difference in the sub-region is For the The phase value corresponding to the minimum difference in the sub-region.

[0051] Among them, The greater the difference between the phase value with the largest difference and the phase value with the smallest difference among the sub-regions, the The phase difference between the sub-regions is also greater.

[0052] S4: Perform region segmentation.

[0053] In response to the phase difference being greater than a preset division threshold, the sub-region is segmented for the next time; otherwise, the region segmentation is stopped to obtain a plurality of sub-regions when the segmentation is finished.

[0054] See also Figure 2 , only the first segmentation will divide the slope interference phase image into four equal sub-regions, and each subsequent segmentation will divide the sub-region that meets the segmentation conditions into four equal sub-regions (although the size of the region obtained by each segmentation is different, as long as the individual regions obtained through regional segmentation are sub-regions), until no sub-regions that meet the segmentation conditions appear, the segmentation is stopped, and the preliminary segmentation result, that is, multiple segmented sub-regions, is obtained.

[0055] S5: Determine the distribution similarity and overall similarity of adjacent sub-regions when the segmentation is terminated.

[0056] It should be noted that, in the subsequent analysis of whether sub-regions (all sub-regions at the end of segmentation) need to be merged, the need for merging can be analyzed based on the similarity between the sub-region and the adjacent sub-regions; some sub-regions with higher similarity are merged, and the remaining sub-regions with lower similarity are not merged to avoid excessive processing and affect the final filtering effect; wherein, the similarity between the sub-region and the adjacent sub-regions here mainly includes the distribution similarity between the two, and the relationship between the phase difference, because by comparing the phase difference between the sub-region and the adjacent sub-region, the similarity between them can be intuitively evaluated, and the distribution similarity can reflect the overall consistency of the two sub-regions in phase value. The combination of the two can further support the decision on sub-region merging.

[0057] For the target subregion and the adjacent subregion among the multiple subregions at the end of segmentation, curve fitting is performed on the two subregions, and the distribution similarity of the two subregions is determined according to the area of ​​the figure formed by the fitting curves of the two subregions.

[0058] Specifically, the distribution similarity satisfies the following relationship:

[0059] ;

[0060] In the formula, The first sub-regions and The distribution similarity of adjacent sub-regions, The first sub-regions and The area of ​​the figure formed by the fitting curves of adjacent sub-regions, The first The fitting curve of the sub-region is The first The sub-region The fitting curves of adjacent sub-regions are and The first sub-regions and The maximum and minimum phase values ​​in adjacent sub-regions, is the natural exponential function, is the definite integral symbol, is the integration variable The differential of is the absolute value symbol.

[0061] in, Indicates the end of the split sub-regions and The area of ​​the figure formed by the fitting curves of adjacent sub-regions is smaller. Sub-regions and The distribution similarity of adjacent sub-regions is higher.

[0062] The overall similarity of the two sub-regions is determined according to the distribution similarity and the phase difference between the two sub-regions.

[0063] Specifically, the overall similarity satisfies the following relationship:

[0064] ;

[0065] In the formula, The first sub-regions and The overall similarity of adjacent sub-regions, The first sub-regions and The distribution similarity of adjacent sub-regions, The first The phase difference corresponding to each sub-region is The first The sub-region The phase difference corresponding to the adjacent sub-regions.

[0066] in, Indicates the end of the split sub-regions and The distribution similarity of adjacent sub-regions is larger, and the smaller the difference between the phase differences of the two sub-regions is, the sub-regions and The larger the number of adjacent sub-regions, the higher the overall similarity.

[0067] S6: Calculate the comprehensive similarity of the sub-regions at the end of segmentation.

[0068] Specifically, the comprehensive similarity of the target sub-region satisfies the following relationship:

[0069] ;

[0070] In the formula, The first The comprehensive similarity of the sub-regions, The first sub-regions and The overall similarity of adjacent sub-regions, The first The number of sub-regions adjacent to a sub-region.

[0071] S7: Perform regional merging.

[0072] Regions are merged according to the size of the comprehensive similarity to obtain multiple final sub-regions.

[0073] Specifically, the regional merging includes:

[0074] In response to the comprehensive similarity being greater than the average of the comprehensive similarities of all sub-regions at the end of segmentation, the target sub-region is merged with the adjacent sub-region with the highest overall similarity; otherwise, the region merging is stopped.

[0075] S8: Filtering is performed according to the size of the standard deviation of the final sub-area to realize the monitoring of slope deformation.

[0076] Specifically, the filtering includes:

[0077] In response to the phase value of the phase point being greater than the standard deviation of the phase value of the final sub-region to which it belongs, the phase point is identified as a pixel gross error point and is removed to obtain a filtered slope interference phase map.

[0078] Specifically, the monitoring of slope deformation includes:

[0079] Phase unwrapping is used to obtain the deformation phase of the filtered slope interference phase image, and then the deformation phase is converted into a deformation map and output, so that the monitoring personnel can intuitively understand the deformation of the slope. At the same time, the warning threshold can also be set according to the deformation analysis results. When the deformation exceeds the threshold, the warning mechanism is triggered (not limited here), reminding the monitoring personnel to take timely measures to complete the slope deformation monitoring based on image recognition.

[0080] Among them, phase unwrapping and deformation phase conversion are both existing technologies and will not be described in detail here.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A slope deformation monitoring method based on image recognition, characterized in that: include: The collected slope interference phase image is divided into regions and filtered using quadtree filtering, including: The slope interference phase map is segmented into multiple sub-regions, and the local differences of the phase points in the sub-regions are obtained. The local differences are used to determine the data deviation of the phase points in the corresponding data set; according to the local differences of the phase points in the sub-regions, the difference in the phase values ​​corresponding to the phase points is determined; the difference in the phase values ​​corresponding to the maximum and minimum difference expression degrees is taken as the phase difference of the sub-region; the local differences satisfy the following relationship: ； In the formula, For the In the sub-area The local difference of the phase points, For the In the sub-area The phase value of the phase point, For the In the sub-area The phase point belongs to the The phase value of the phase point, For the The number of phase points in the data set to which the phase point belongs, is the absolute value symbol; The degree of difference expression satisfies the following relationship: ； In the formula, For the Phase value in sub-region The degree of difference in performance, For the In the sub-area The phase value is The local difference of the phase point, For the The phase value in the sub-region is The number of phase points; In response to the phase difference being greater than a preset division threshold, performing the next region division on the sub-region, otherwise, stopping the region division to obtain a plurality of sub-regions when the division ends; For the target sub-region and the adjacent sub-region in the multiple sub-regions at the end of segmentation, curve fitting is performed on the two sub-regions, and the distribution similarity of the two sub-regions is determined according to the area of ​​the figure formed by the fitting curves of the two sub-regions; the overall similarity of the two sub-regions is determined according to the distribution similarity and the phase difference of the two sub-regions; the overall similarity of the target sub-region and all adjacent sub-regions is weighted averaged to obtain the comprehensive similarity of the target sub-region; the distribution similarity satisfies the following relationship: ； In the formula, The first sub-regions and The distribution similarity of adjacent sub-regions, The first sub-regions and The area of ​​the figure formed by the fitting curves of adjacent sub-regions, The first The fitting curve of the sub-region is The first The sub-region The fitting curves of adjacent sub-regions are and The first sub-regions and The maximum and minimum phase values ​​in adjacent sub-regions, is the natural exponential function, is the definite integral symbol, is the integration variable The differential of is the absolute value symbol; The overall similarity satisfies the following relationship: ； In the formula, The first sub-regions and The overall similarity of adjacent sub-regions, The first sub-regions and The distribution similarity of adjacent sub-regions, The first The phase difference corresponding to each sub-region is The first The sub-region The phase difference corresponding to adjacent sub-regions; Merge regions according to the size of comprehensive similarity to obtain multiple final sub-regions; Filtering is performed according to the size of the standard deviation of the final sub-area to achieve slope deformation monitoring.

2. The slope deformation monitoring method based on image recognition according to claim 1 is characterized in that: The regional merging comprises: In response to the comprehensive similarity being greater than the average of the comprehensive similarities of all sub-regions at the end of segmentation, the target sub-region is merged with the adjacent sub-region with the highest overall similarity; otherwise, the region merging is stopped.

3. The slope deformation monitoring method based on image recognition according to claim 1 is characterized in that: The filtering comprises: In response to the phase value of the phase point being greater than the standard deviation of the phase value of the final sub-region to which it belongs, the phase point is identified as a pixel gross error point and is removed to obtain a filtered slope interference phase map.

4. The slope deformation monitoring method based on image recognition according to claim 3 is characterized in that: The monitoring of slope deformation is achieved by: Phase unwrapping is used to obtain the deformation phase of the filtered slope interference phase image, and then the deformation phase is converted into a deformation map and output to complete the slope deformation monitoring based on image recognition.

Images