Edge two-phase interface completion method based on time-series images

Abstract

The application discloses a kind of edge two-phase interface completion methods based on time sequence image, after being converted into three-dimensional matrix to be handled time sequence image, the incomplete discrete phase of any frame image flow outlet boundary of three-dimensional matrix is identified, and the data of current frame is updated after being expanded and completed based on its previous frame image, based on the three-dimensional matrix updated, the incomplete discrete phase of any frame image flow outlet boundary of three-dimensional matrix is identified again, and the data of current frame is updated again after being expanded and completed based on its next frame image for the second time, and the edge two-phase interface completion is realized.The application is directed to the time sequence image of two-phase flow, and the incomplete phase interface located at the edge of image is completed, and the expansion completion algorithm not dependent on neural network is avoided, and the inconvenience caused by preparing a large number of training data in advance is avoided.

Description

Technical Field

[0001] This invention relates to a technique in the field of image processing, specifically a method for edge two-phase interface completion based on time-series images. Background Technology

[0002] Two-phase flow is widely used in industrial equipment in fields such as refrigeration, petroleum, and energy. To improve the design performance of related industrial equipment, researchers often use wire mesh measurement sensors or high-speed cameras to visualize and measure two-phase flow. Based on the obtained binary time-series images of the two-phase flow, they analyze the quantity, size, and shape of the discrete phases in the flow, thereby supporting the characteristic analysis of the two-phase flow. Therefore, accurately and reliably obtaining information on the quantity, size, and shape of discrete phases from the time-series images of two-phase flow is crucial for experimental research on two-phase flow. However, due to the limitation of the measurement field of view, when the discrete phase flows near the edge of the measurement area, some phase interfaces will be located outside the measurement area. Therefore, incomplete phase interfaces will inevitably appear at the edges in the obtained images. If statistical analysis is performed directly on the measured two-phase images, the size of the discrete phases located at the edges will be significantly underestimated, and incorrect discrete phase shapes will be obtained. Therefore, it is necessary to use algorithms to complete the phase interfaces of the discrete phases at the edges to avoid large errors in data analysis. Summary of the Invention

[0003] This invention addresses the shortcomings of existing technologies, such as their difficulty in completing incomplete phase interfaces at the edges of binary images of two-phase flows, the complexity and high processing costs of such methods, and the current lack of outward extension completion technology for incomplete interfaces at image edges. It proposes a two-phase interface completion method based on time-series images. This method completes incomplete phase interfaces at the image edges of time-series images of two-phase flows without relying on neural network extension completion algorithms, thus avoiding the inconvenience of preparing large amounts of training data in advance.

[0004] This invention is achieved through the following technical solution:

[0005] This invention relates to a method for edge two-phase interface completion based on time-series images. After converting the time-series image to be processed into a three-dimensional matrix, the method identifies the incomplete discrete phase at the flow outlet boundary of any frame of the three-dimensional matrix, performs expansion and completion based on the previous frame, and updates the data of the current frame. Based on the updated three-dimensional matrix, the method identifies the incomplete discrete phase at the flow outlet boundary of any frame of the three-dimensional matrix again, performs secondary expansion and completion based on the next frame, and updates the data of the current frame again, thereby achieving edge two-phase interface completion.

[0006] The extended completion includes:

[0007] Step a: Extend the coverage area of ​​the k-th incomplete discrete phase circumferentially to establish the search area Ψ. search=([i s，min，t，k i s,max,t,k ],[j s,min，t,k j s，max,t,k ]), where: i s,min，t,k =max(i min,t,k -Φd t,k ,1), j s，min,t,k =max(j min,t,k -Φd t，k ,1), Φ is the search expansion coefficient, which can be determined according to the actual situation. Let be the maximum value of the coordinates in the image along the i-th direction in matrix P. This represents the maximum coordinate value in the j-direction of the image within matrix P, and it changes continuously with each calculation loop. Therefore, it is used in each call... The current size of matrix P needs to be checked in all cases.

[0008] Step b, for the mask matrix M t,k Within the coordinate range Ψ of the covered area cover The image and matrix P within the search area Ψ of frame t-1 or frame t+1. search Edge detection is performed on the image within the image using the Canny or Sobel algorithm, and the output results are cross-correlated. The position of the maximum value in the cross-correlation result is recorded (i). r，t，k j r,t，k ).

[0009] Step c: Use the region seed growth algorithm to (i r，t，k -i s，max，t,k +i s，min，t，k j r，t，k -j s，max，t,k +j s，min,t，k Using the seed coordinates, identify the discrete phase regions in the copy matrix P where the (t-1)th or (t+1)th frame contains the seed coordinates, and obtain the mask matrix of the discrete phase. or

[0010] Step d, the location of the maximum value in the cross-correlation calculation results from step c (i r，t，k j r，t,k ), M t,k Translate the discrete phase region to the mask matrix or The corresponding pixel value is marked as 2 to form a matching mask. or

[0011] Step e: Match the mask or The process involves iterating through the map and filtering for pixels with a value of 1 that satisfy j1 > max(j2). This filters and determines the region to be filled in for the k-th incomplete discrete phase. Here, j1 represents the coordinates of the pixel with a value of 1 in the matching mask along the j-direction, and j2 represents the coordinates of the pixel with a value of 2 in the matching mask along the j-direction. After changing the pixel values ​​of the filtered pixels to 2, the matching mask is then... Change the pixel value of the point with a pixel value of 1 to 0, and then change the pixel value of the point with a pixel value of 2 to 1 to obtain the optimized mask. or

[0012] Step f: Optimize the mask or Reverse translation to M t,k The extended and completed mask of the k-th incomplete discrete phase boundary of the t-th frame of the three-dimensional matrix B1 is obtained. Calculate k from 1 to N in a loop t,k The obtained extended and completed masks are superimposed onto the t-th frame image of the 3D matrix B1 to obtain the t-th frame extended and completed image.

[0013] This invention relates to a system for implementing the above method, comprising: a data preprocessing unit, an expansion and completion unit, and a data integration unit, wherein: the data preprocessing unit performs data format conversion and incomplete discrete phase marking based on the input time-series image information to obtain three-dimensional matrix data containing incomplete discrete phase marking information; the expansion and completion unit expands and completes the incomplete discrete phases marked by the matrix based on the three-dimensional matrix data output by the data preprocessing unit to obtain three-dimensional matrix data containing the expansion and completion results; and the data integration unit performs data format conversion based on the three-dimensional matrix data output by the expansion and completion unit and outputs the expanded and completed time-series image information. Technical effect

[0014] Compared with existing technologies, this invention does not rely on neural networks, which can avoid the inconvenience of preparing a large amount of training data in advance. At the same time, it can effectively complete the incomplete discrete phase boundary structure of the image edge. It only relies on the algorithm to realize the edge completion of time series images applicable to various two-phase flows. It has low requirements for computer performance and fast calculation speed. Attached Figure Description

[0015] Figure 1 This is a flowchart of the present invention;

[0016] Figure 2 This is a diagram showing the effect after the incomplete interface at the edges of the example is expanded and completed;

[0017] Figure 3 This is a comparison chart of the completed results and actual data for the example;

[0018] Figure 4 This is a quantitative comparison and evaluation chart of the completed results and real data for the example. Detailed Implementation

[0019] like Figure 1 As shown, this embodiment relates to a method for edge two-phase interface completion based on time-series images, including:

[0020] Step 1: Read and store the binarized time series image of the two-phase flow as a three-dimensional matrix B1 = {δ} i，j，t |1≤i≤N i ,1≤j≤N j ,1≤t≤N t}, then create a replica matrix P = B1, where: δ i,j,t This refers to the image pixel value at position (i, j) in frame t; j is the position coordinate of the two-phase flow direction, j = 1 indicates that the pixel is located at the boundary of the two-phase flow inlet in the image, j = N j This indicates that the pixel is located at the two-phase flow outlet boundary of the image; i represents the position coordinates perpendicular to the flow direction. N i ×N j This represents the pixel dimensions of the image frame. δ i,j,t =1 indicates that the image at that location is in discrete phase, δ i,j,t =0 indicates that the image at that point is a continuous phase.

[0021] Step 2: Identify the complete discrete phase and the incomplete discrete phase located at the exit boundary in the t-th frame image of the 3D matrix B1. Obtain the coverage area and feature length of the incomplete discrete phase corresponding to the mask matrix machine of the incomplete discrete phase at the exit boundary. Specifically, iterate through the pixel values ​​of the t-th frame image of the 3D matrix B1 and filter δ i,j,t =1 and j=N j The pixels are used to obtain the set of discrete phase pixels located at the exit boundary. Using a region seeding algorithm, the discrete phase regions containing each pixel are identified sequentially, with the pixels in B2 as seed coordinates, to obtain the mask matrix M of the incomplete discrete phase at the exit boundary of the t-th frame image. t，k Based on the positions of non-zero pixels in the mask matrix, the coordinate range Ψ of the coverage area corresponding to the incomplete discrete phase is further obtained. cover =([i min,t,k i max,k ],[j min,t,k N j The centroid coordinates (i) of the incomplete discrete phase are calculated. c，t，k j c，t，k and feature length Where: k = 1, ..., N t，k Nt，k Let δ′ be the number of incomplete discrete phases at the exit boundary of the t-th frame image. i，j，t ∈M t,k .

[0022] The mask matrix M t,k In the diagram, the pixel value of the k-th incomplete discrete phase coverage area is 1, while the pixel value of the remaining areas is 0.

[0023] Step 3: Based on the replica matrix P, extend and complete the incomplete discrete phase coverage area of ​​the t-th frame obtained in Step 2 to obtain the extended and completed image of the t-th frame, specifically including:

[0024] 3.1 Extend the coverage area of ​​the k-th incomplete discrete phase circumferentially to establish the search area Ψ. search =([i s，min，t，k i s，max，t，k ],[j s，min，t，k j s，max，t，k ]), where: i s，min，t，k =max(i min，t，k -Φd t，k ,1), j s，min，t，k =max(j min，t，k -Φd t，k ,1), Φ is the search expansion coefficient, which can be determined according to the actual situation. Let be the maximum value of the coordinates in the image along the i-th direction in matrix P. This represents the maximum coordinate value in the j-direction of the image within matrix P, and it changes continuously with each calculation loop. Therefore, it is used in each call... The current size of matrix P needs to be checked in all cases.

[0025] 3.2 Mask matrix M t，k Within the coordinate range Ψ of the covered area cover The search area Ψ of the image and matrix P in frame t-1 search Edge detection is performed on the image within the image using the Canny or Sobel algorithm, and the output results are cross-correlated. The position of the maximum value in the cross-correlation result is recorded (i). r，t，k j r，t，k ).

[0026] 3.3 Using the regional seed growth algorithm, with (i r，t，k -i s，max，t，k +i s，min，t，k j r，t，k -j s，max，t，k +j s，min，t，kUsing seed coordinates, identify the discrete phase region in the (t-1)th frame of the copy matrix P that contains the seed coordinates, and obtain the mask matrix of the discrete phase.

[0027] 3.4 The location of the maximum value in the cross-correlation calculation results in step 3.3 (i r,t，k j r，t，k ), M t，k Translate the discrete phase region to the mask matrix The corresponding pixel value is marked as 2 to form a matching mask.

[0028] 3.5 Matching Mask The process involves iterating through the map and filtering for pixels with a value of 1 that satisfy j1 > max(j2). This filters and determines the region to be filled in for the k-th incomplete discrete phase. Here, j1 represents the coordinates of the pixel with a value of 1 in the matching mask along the j-direction, and j2 represents the coordinates of the pixel with a value of 2 in the matching mask along the j-direction. After changing the pixel values ​​of the filtered pixels to 2, the matching mask is then... Change the pixel value of the point with a pixel value of 1 to 0, and then change the pixel value of the point with a pixel value of 2 to 1 to obtain the optimized mask.

[0029] The matching mask The image contains three types of pixels: pixels with values ​​of 0, 1, and 2. Let the pixels with a value of 1 be denoted as... represent China was not M t,k Discrete phase region covered by discrete phase; Pixels with a pixel value of 2 are denoted as... Representing M t，k The region after discrete phase translation.

[0030] Version 3.6 will optimize the mask. Reverse translation to M t,k The extended and completed mask of the k-th incomplete discrete phase boundary of the t-th frame of the three-dimensional matrix B1 is obtained. Calculate k from 1 to N in a loop t，k The obtained extended and completed masks are superimposed onto the t-th frame image of the 3D matrix B1 to obtain the t-th frame image of the exit boundary after extension and completion.

[0031] Step 4: Replace the t-th frame data in the replica matrix P with the t-th frame image data obtained in Step 3 after the export boundary extension and completion. Specifically, based on... The size of matrix P is adjusted, and the newly added region of the matrix has a pixel value of 0. The maximum coordinate value in the j-th direction of the t-th frame image after the export boundary extension and completion is determined. Subsequently, the t-th frame image in the corrected matrix P is replaced with the t-th frame image after the export boundary extension and completion, thus completing the matrix P update. Steps 3 and 4 are executed iteratively until the Nth frame image is obtained. t Frame, i.e., t=N t This ensures that each frame in the replica matrix P is an image after the export boundary extension and completion.

[0032] Step 5: Create the updated time-series image data matrix B new =P, identify the matrix B new The complete discrete phase and the incomplete discrete phase located at the entrance boundary in the t-th frame image are obtained. Specifically, the mask matrix of the incomplete discrete phase at the entrance boundary, the coverage area of ​​the corresponding incomplete discrete phase, and the feature length are obtained. This is achieved by updating the time-series image data matrix B. new Iterate through the pixel values ​​of the t-th frame image and filter δ. i，j，t The set of discrete phase pixels located at the exit boundary is obtained by considering pixels with j=1 and j=1. Using a region seeding algorithm, the discrete phase regions containing each pixel are identified sequentially, with the pixels contained in B2′ as seed coordinates, to obtain the updated mask matrix M of the incomplete discrete phase at the exit boundary of the t-th frame image. t,k Based on the positions of non-zero pixels in the mask matrix, the coordinate range Ψ of the updated coverage area corresponding to the incomplete discrete phase is further obtained. cover ′=([i min,t，k i max，k ],[j min,t,k N j And calculate the updated centroid coordinates (i) of the incomplete discrete phase. c，t，k ′,j c，t，k ′) and update feature length Where: k = 1, ..., N t，k N t，k Let δ′ be the number of incomplete discrete phases at the exit boundary of the t-th frame image. i，j，t ∈M t，k .

[0033] Step 6: Based on the matrix P obtained in Step 4, extend and complete the incomplete discrete phase coverage area of ​​the t-th frame image obtained in Step 5 to obtain the extended and completed image of the t-th frame, specifically including:

[0034] 6.1 Extend the coverage area of ​​the k-th incomplete discrete phase circumferentially to establish the search area Ψ. search =([i s，min，t，k i s，max，t，k ],[j s，min，t，k js，max，t，k ]), where: i s，min，t，k =max(i min，t，k -Φd t，k ,1), j s，min,t,k =max(j min，t，k -Φd t，k ,1), Φ is the search expansion coefficient, which can be determined according to the actual situation. Let be the maximum value of the coordinates in the image along the i-th direction in matrix P. This represents the maximum coordinate value in the j-direction of the image within matrix P, and it changes continuously with each calculation loop. Therefore, it is used in each call... The current size of matrix P needs to be checked in all cases.

[0035] 6.2 Mask Matrix M t，k Within the coordinate range Ψ of the covered area cover The search area Ψ of the image and matrix P in frame t+1. search Edge detection is performed on the image within the image using the Canny or Sobel algorithm, and the output results are cross-correlated. The position of the maximum value in the cross-correlation result is recorded (i). r，t，k j r，t，k ).

[0036] 6.3 Using the regional seed growth algorithm, with (i r，t，k -i s，max，t，k +i s，min，t，k j r，t，k -j s，max，t，k +J s，min，t，k Using seed coordinates, identify the discrete phase region in the (t+1)th frame of the copy matrix P that contains the seed coordinates, and obtain the mask matrix of the discrete phase.

[0037] 6.4 The location of the maximum value in the cross-correlation calculation results based on step 3.3 (i r，t，k j r，t，k ), M t，k Translate the discrete phase region to the mask matrix The corresponding pixel value is marked as 2 to form a matching mask.

[0038] 6.5 Matching Mask traverse, screen to obtain pixel points with a pixel value of 1 that satisfy j1<min(j2), that is, screen and determine the region to be complemented of the k-th incomplete discrete phase, wherein j1 refers to the coordinate in the j-direction of the pixel point with a pixel value of 1 in the matching mask, and j2 refers to the coordinate in the j-direction of the pixel point with a pixel value of 2 in the matching mask. After changing the pixel values of the screened pixel points to 2, the matching mask changes the pixel values of points with a pixel value of 1 to 0, and then changes the pixel values of points with a pixel value of 2 to 1, so as to obtain an optimized mask

[0039] the matching mask described in contains three types of pixel points, which are pixel points with pixel values of 0, 1 and 2 respectively, wherein: pixel points with a pixel value of 1 are recorded as representing discrete phase regions not covered by M t，k discrete phase; pixel points with a pixel value of 2 are recorded as representing the region after translation of M t，k discrete phase.

[0040] 6.6 Translate the optimized mask reverse-translate to M t，k to obtain an expanded and complemented mask for the k-th incomplete discrete phase boundary of the t-th frame in the three-dimensional matrix B1 circularly calculate k from 1 to N t，k , superimpose each obtained expanded and complemented mask onto the t-th frame image of the three-dimensional matrix B1, to obtain the t-th frame image after the expansion and complementation of the t-th inlet boundary.

[0041] Step 7, use the t-th frame image data after expansion and complementation to replace the t-th frame data in the matrix P. This step is used to update the size and data of the matrix P output in step 4, so that each frame of image in the matrix P is an image after expansion and complementation of the inlet and outlet boundaries, which specifically includes: based on correct the size of the matrix P, the pixel value of the newly added region of the matrix is 0, wherein is the maximum coordinate value in the j-direction of the t-th frame image after expansion and complementation of the outlet boundary. Then, the t-th frame image after expansion and complementation of the outlet boundary is used to replace the t-th frame image in the corrected matrix P, so as to complete the update of the matrix P. Circularly execute step 5 and step 6 until the first frame, that is, t=1, so that each frame of image in the copy matrix P is an image after expansion and complementation of the inlet and outlet boundaries.

[0042] Through a specific actual experiment, for the binarized sequential images of gas-liquid two-phase flow with a resolution of 320×320 and 15000 frames, on a computer with 32G memory, an Intel Core i7 processor and no independent graphics card, the above method is run with an expansion coefficient Φ=2, and the result is obtained as shown in Figure 2(Right) shows a two-phase flow time series image with complete discrete phase boundaries.

[0043] like Figure 2 As shown on the left, due to limitations of the actual measurement range, the original image exhibits incomplete gas phase structures (i.e., incomplete discrete phase structures) at both the inlet and outlet edges of the flow. The method described in this invention effectively expands and fills in the incomplete gas phase boundaries, providing a more accurate reflection of the actual phase boundary distribution. To further demonstrate the accuracy of this invention, the binarized time-series image of the vapor-liquid two-phase flow is artificially divided into two halves along the flow direction. Only half of the time-series image is input into the method described in this invention for completion. The completed gas phase image is then compared and verified with the actual complete image.

[0044] like Figure 3 As shown, the distribution of the phase interface after processing half of the image data using this method is compared with the distribution of the complete actual phase interface at different times. It can be found that the boundary after the extension and completion of the present invention matches the actual boundary shape well and can accurately reflect the actual distribution of the incomplete phase boundary of the image edge.

[0045] like Figure 4 As shown, the relative error between the projected perimeter and projected area of ​​the gas phase region after the expansion and completion of the present invention and the actual value can be found that the error after processing by the present invention is within ±10%, and will decrease as the size of the gas phase structure increases.

[0046] In summary, this method can complete the incomplete discrete phase boundaries of time-series images without relying on neural network methods and training data, and the completion results are effective, reliable, and in good agreement with the real situation.

[0047] The above-described specific implementations can be partially adjusted by those skilled in the art in different ways without departing from the principles and purpose of the present invention. The scope of protection of the present invention is defined by the claims and is not limited to the above-described specific implementations. All implementation schemes within the scope of the claims are bound by the present invention.

Claims

1. A method for edge two-phase interface completion based on time-series images, characterized in that, After converting the time series image to be processed into a three-dimensional matrix, the incomplete discrete phase at the flow outlet boundary of any frame of the three-dimensional matrix is ​​identified and expanded and completed based on the previous frame image, and the data of the current frame is updated. Based on the updated three-dimensional matrix, the incomplete discrete phase at the flow inlet boundary of any frame of the three-dimensional matrix is ​​identified and expanded and completed based on the next frame image, and the data of the current frame is updated again, so as to realize the completion of the two-phase interface at the edge. The extended completion includes: Step a: Extend the coverage area of ​​the k-th incomplete discrete phase circumferentially to establish the search area. ,in: , , , , This is a search expansion factor, which can be determined according to the actual situation. Let be the maximum value of the coordinates in the image along the i-th direction in matrix P. This represents the maximum coordinate value in the j-direction of the image within matrix P, and it changes continuously with each calculation loop. Therefore, it is used in each call... The current size of matrix P needs to be checked in all cases; Step b, Apply mask matrix Within the coordinate range of the coverage area The image and matrix P within the (t-1)th frame or the (t+1)th frame Edge detection is performed on the image within the image using the Canny or Sobel algorithm, and the output results are cross-correlated. The location of the maximum value in the cross-correlation result is recorded. ); Step c: Utilize the region seed growth algorithm to ( Using the seed coordinates, identify the discrete phase regions in the copy matrix P where the (t-1)th or (t+1)th frame contains the seed coordinates, and obtain the mask matrix of the discrete phase. or ; Step d, the location of the maximum value in the cross-correlation calculation results from step c ( ),Will Translate the discrete phase region to the mask matrix or The corresponding pixel value is marked as 2 to form a matching mask. or ; Step e: Match the mask or Perform a traversal and filter to find pixels with a value of 1 that satisfy the following conditions: The pixels are selected to fill in the k-th incomplete discrete phase. Here, j1 refers to the coordinates of a pixel with a value of 1 in the matching mask along the j-direction, and j2 refers to the coordinates of a pixel with a value of 2 in the matching mask along the j-direction. After changing the pixel values ​​of the selected pixels to 2, the matching mask is... Change the pixel value of the point with a pixel value of 1 to 0, and then change the pixel value of the point with a pixel value of 2 to 1 to obtain the optimized mask. or , Step f: Optimize the mask or Reverse translation to To obtain a three-dimensional matrix Extended and completed mask of the k-th incomplete discrete phase boundary in frame t. Calculate k iteratively from 1 to ... The obtained extended and completed masks are then overlaid onto the 3D matrix. From the t-th frame image, obtain the expanded and completed t-th frame image; The aforementioned re-identification refers to the creation of an updated temporal image data matrix. Identify the matrix The complete discrete phase and the incomplete discrete phase located at the entrance boundary in the t-th frame image are obtained. Specifically, the mask matrix of the incomplete discrete phase at the entrance boundary, the coverage area of ​​the corresponding incomplete discrete phase, and the feature length are obtained. This involves updating the temporal image data matrix... Iterate through the pixel values ​​of the t-th frame image and filter them. and The pixels are used to obtain the set of discrete phase pixels located at the entrance boundary. Using the regional seed growth algorithm, sequentially... The contained pixels serve as seed coordinates. The discrete phase regions containing these pixels are identified to obtain the updated mask matrix for the incomplete discrete phase at the entry boundary of the t-th frame image. Based on the positions of non-zero pixels in the mask matrix, the coordinate range of the updated coverage area corresponding to the incomplete discrete phase is further obtained. , And calculate the updated centroid coordinates of the incomplete discrete phase. and update feature length ,in: , Let be the number of incomplete discrete phases at the entry boundary of the t-th frame image. ; The aforementioned update, which involves replacing the t-th frame data in matrix P with the expanded and completed t-th frame image data, is used to update the size and data of matrix P output in step 4, ensuring that each frame image within matrix P is an expanded and completed image of the entrance and its boundary. Specifically, this includes: based on... The size of matrix P is adjusted, and the newly added region of the matrix has a pixel value of 0. Find the maximum coordinate value in direction j of the t-th frame image after the entry boundary extension and completion. Then, replace the t-th frame image in the corrected matrix P with the t-th frame image after the entry boundary extension and completion, thus completing the matrix P update. Repeat steps 5 and 6 until the t-th frame image is completed. Frame, i.e., t= This ensures that each frame in the replica matrix P is an image of the entry point and the entry point boundary after expansion and completion.

2. The edge two-phase interface completion method based on time-series images according to claim 1, characterized in that, The aforementioned three-dimensional matrix specifically refers to: reading and storing the binarized time-series image of the two-phase flow as a three-dimensional matrix. Then create a replica matrix. ,in: This refers to the image pixel value at position (i, j) in frame t; j is the position coordinate of the two-phase flow direction, j=1 indicates that the pixel is located at the boundary of the two-phase flow inlet in the image, j= This indicates that the pixel is located at the two-phase flow outlet boundary of the image; i represents the position coordinates perpendicular to the flow direction. This refers to the pixel specifications of the image frame. This indicates that the image at this location is in discrete phase. This indicates that the image at that location is a continuous phase.

3. The edge two-phase interface completion method based on time-series images according to claim 1, characterized in that, The aforementioned recognition refers to the recognition of a three-dimensional matrix. The complete discrete phase and the incomplete discrete phase located at the exit boundary in the t-th frame image are used to obtain the mask matrix of the incomplete discrete phase at the exit boundary and the corresponding coverage area and feature length of the incomplete discrete phase. Specifically, for the three-dimensional matrix... Iterate through the pixel values ​​of the t-th frame image and filter them. and The pixels are used to obtain the set of discrete phase pixels located at the exit boundary. Using the regional seed growth algorithm, sequentially... The contained pixels serve as seed coordinates. The discrete phase regions containing these pixels are identified to obtain the mask matrix of the incomplete discrete phase at the exit boundary of the t-th frame image. Based on the positions of non-zero pixels in the mask matrix, the coordinate range of the coverage area corresponding to the incomplete discrete phase is further obtained. , And calculate the centroid coordinates of the incomplete discrete phase. and feature length ,in: , Let be the number of incomplete discrete phases at the exit boundary of the t-th frame image. .

4. The edge two-phase interface completion method based on time-series images according to claim 3, characterized in that, The mask matrix In the diagram, the pixel value of the k-th incomplete discrete phase coverage area is 1, while the pixel value of the remaining areas is 0.

5. The edge two-phase interface completion method based on time-series images according to claim 1, characterized in that, The matching mask The image contains three types of pixels: pixels with values ​​of 0, 1, and 2. Let the pixels with a value of 1 be denoted as... ,represent China was not Discrete phase region covered by discrete phase; Pixels with a pixel value of 2 are denoted as... ,represent The region after discrete phase translation.

6. The edge two-phase interface completion method based on time-series images according to claim 1, characterized in that, The aforementioned updating of the current frame data refers to replacing the t-th frame data in the replica matrix P with the t-th frame image data after the exit boundary extension and completion. Specifically, it is based on... The size of matrix P is adjusted, and the newly added region of the matrix has a pixel value of 0. Find the maximum coordinate value in the j-th direction of the t-th frame image after the export boundary extension and completion. Then, replace the t-th frame image in the corrected matrix P with the t-th frame image after the export boundary extension and completion, thus completing the matrix P update. Repeat steps 3 and 4 until the t-th frame image is completed. Frame, i.e., t= This ensures that each frame in the replica matrix P is an image after the export boundary extension and completion.

7. A two-phase edge interface completion system based on a time-series image, implementing the method of any one of claims 1-6, characterized in that, include: The system comprises a data preprocessing unit, an extension and completion unit, and a data integration unit. The data preprocessing unit performs data format conversion and incomplete discrete phase labeling based on the input time-series image information, obtaining three-dimensional matrix data containing incomplete discrete phase labeling information. The extension and completion unit extends and completes the incomplete discrete phases labeled in the matrix based on the three-dimensional matrix data output by the data preprocessing unit, obtaining three-dimensional matrix data containing the extension and completion results. The data integration unit performs data format conversion based on the three-dimensional matrix data output by the extension and completion unit, outputting the extended and completed time-series image information.

Images