A target pellet interferogram unwrapping method based on edge recognition and phase stratification

Abstract

The present invention discloses a target-pill interferogram unwrapping method based on edge recognition and phase stratification, comprising: (1) performing edge extraction on the input target-pill interferogram to extract the edges between each phase layer; (2) selecting the starting point of each phase layer to obtain a starting point sequence S(n) of each phase layer, and performing denoising processing; (3) traversing the selected starting sequence points S(n), performing a breadth-first search on each starting point, identifying the entire phase layer in which it is located, and recording the phase layer number of the searched pixel; (4) performing an overall phase shift on the identified phase layer to correct the residual jump points in the phase matrix, thereby obtaining an unwrapped phase φ. Utilizing the present invention, the target-pill interferogram image can be unwrapped quickly and accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of optical precision detection, and in particular relates to a target pellet interference pattern unwrapping method based on edge recognition and phase stratification. Background Art

[0002] Inertial confinement fusion (ICF) uses multiple laser beams to bombard a fuel target pellet, producing controlled thermonuclear fusion and releasing huge amounts of energy. It is one of the two mainstream solutions for achieving controlled nuclear fusion. In order to achieve fusion conditions, the parameters of the spherical shell and fuel ice layer of the target pellet, which serves as the convergence center of the multiple laser beams, must be strictly controlled. Among them, precise detection of the refractive index and thickness of each layer is crucial. Since the diameter of the target pellet is only a few millimeters, small changes in the thickness of its ice layer are difficult to characterize using conventional measurement methods. Interferometry, with its non-destructive detection and high resolution, has become an important means of accurately measuring the ice layer of the target pellet.

[0003] Current research on pellet interferometry utilizes phase-shifting methods to process multiple interferograms, requiring high instrument precision and being susceptible to environmental influences. Therefore, phase extraction algorithms for single interferograms have become a key area of ​​research. Spatial-domain phase measurement methods are suitable for processing single interferograms and can achieve high accuracy, but they inevitably involve arctangent calculations, causing the resulting phase result to be wrapped between (-π, π). Therefore, unwrapping algorithms are crucial to obtaining the true phase of the measured object.

[0004] Conventional unwrapping methods are mainly divided into path-following algorithms and path-independent algorithms. Path-following algorithms traverse all points in an image according to a certain rule and integrate the phase gradient along the path to achieve phase unwrapping. Since various noise and defects in real images can cause cumulative errors in the path integral, selecting an appropriate path is crucial for this unwrapping method. Richard M. Goldstein et al. proposed a branch-cutting method, which allows the path to avoid all positive and negative residual points in the wrapping. Souza et al. improved this method by using jump charges, which is fast and memory-efficient, but it is still prone to branch-cutting loops, causing unwrapping failure. Xu et al. proposed a quality-graph-guided method, which uses a quality graph to gradually diffuse unwrapping from high-quality points to low-quality points. Liang Man et al. subsequently improved this method by using wavelet-based profile analysis. This method is fast but relies on a high-quality quality graph. Flynn et al. proposed a minimum discontinuity algorithm, which uses a specific path structure graph to guide phase unwrapping to minimize the sum of discontinuities. This method does not require preliminary work such as a quality graph or branch-cutting, but the search for discontinuities leads to high computational complexity and low efficiency. These algorithms all utilize some degree of judgment and optimization for path selection, but they are not effective in unwrapping complex phase conditions or extremely high levels of noise. Path-independent algorithms, which do not rely on specific paths, are better suited for real-world scenarios. The most representative algorithm is the least squares method. This method minimizes the difference between the partial derivatives of the wrapped phase and the actual phase, making it well suited for complex noise scenarios. However, it also smoothes the phase points surrounding the noise to some extent. Summary of the Invention

[0005] The present invention provides a target-pill interferogram unwrapping method based on edge recognition and phase stratification, which can quickly and accurately unwrap the target-pill interferogram image.

[0006] A target pellet interferogram unwrapping method based on edge recognition and phase stratification, comprising:

[0007] (1) Perform edge extraction on the input target pellet interference wrapping pattern to extract the edges between each phase layer;

[0008] (2) Select the starting point of each phase layer to obtain the starting point sequence S(n) of each phase layer and perform denoising;

[0009] (3) Traverse the selected starting sequence points S(n), perform breadth-first search on each starting point, identify the entire phase layer where it is located, and record the phase layer number of the searched pixel;

[0010] (4) Perform an overall phase shift on the identified phase layer to correct the remaining error points in the phase matrix, thereby obtaining the unwrapped phase φ.

[0011] Furthermore, in step (1), the size of the target pellet interference wrapping pattern is M×N, is the wrapping phase of the pixel coordinate (x, y), and its value range is The value range of i and j is 0≤x≤M-1, 0≤y≤N-1.

[0012] Furthermore, in step (1), when performing edge extraction, the LOG operator is used as the edge extraction operator. The LOG operator first performs a Gaussian low-pass filter on the image, which is expressed as:

[0013]

[0014] Where f(x, y) is the image and G(x, y) is the Gaussian function, which is expressed as:

[0015]

[0016] Where σ is the standard deviation. Since this is a linear system, the order of convolution and differentiation calculation in equation (1) is swapped to obtain:

[0017]

[0018] Equation (3) shows that the Gaussian operator is differentiated first and then convolved with the image f(x, y); therefore, the derivative of Equation (2) is calculated as:

[0019]

[0020] Formula (4) is called the LOG operator.

[0021] Furthermore, the specific process of step (2) is:

[0022] For the target pellet interference wrapping diagram, the center of the target pellet is selected as the starting point S(0) of the zeroth phase layer. A line is drawn outward from this point, which intersects with the edge of each layer at points P(1), P(2)... The midpoint of every two adjacent P points is selected as the starting point S(i) of the i-th layer; thus, the starting point sequence coordinates of each phase layer are:

[0023]

[0024] Finally, the starting point sequence S(n) of each phase layer is obtained.

[0025] Furthermore, in step (2), when performing denoising, a connected graph algorithm is used to remove connected graphs with points less than a set threshold, while retaining coherent edges.

[0026] Furthermore, in step (4), to correct the remaining erroneous points in the phase matrix, it is necessary to compare the phases of all pixels in the entire image with those of the surrounding points and compensate for the phase difference. The specific formula is:

[0027]

[0028] Where R is the rounding function, and M is the median phase value within a certain range around the point (x, y).

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. This invention innovatively combines wrapper map processing with the edge recognition algorithm in the machine vision algorithm, and uses the fault edges of the wrapper map phase grayscale image to stratify the phase, effectively compensating for the cumulative error of the path tracking algorithm and the shortcomings of the least squares method in smoothing the true phase, thereby accurately restoring the true phase value of the wrapper map.

[0031] 2. The algorithm of the present invention has moderate complexity and high execution speed. It can not only quickly and accurately unwrap the interferogram image of the target pellet, but also quickly and accurately process the concentric wrapped phase of other complex shapes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a flow chart of a target pellet interferogram unwrapping method based on edge recognition and phase stratification according to the present invention;

[0033] Figure 2 This is an edge recognition graph before and after denoising of the connected graph in an embodiment of the present invention;

[0034] Figure 3 A starting point sequence diagram selected in an embodiment of the present invention;

[0035] Figure 4 A flow chart of the present invention for identifying all phase layers based on a starting point sequence;

[0036] Figure 5 This is a phase diagram before and after the residual jump point is eliminated in an embodiment of the present invention;

[0037] Figure 6 This is a result diagram of the application of the present invention to a certain target pellet interference pattern. DETAILED DESCRIPTION

[0038] The present invention will be described in further detail below with reference to the accompanying drawings and examples. It should be noted that the following examples are intended to facilitate understanding of the present invention and do not have any limiting effect on the present invention.

[0039] For the package of the target pellet interference diagram, let the package diagram size be M×N. is the wrapping phase of the pixel coordinate (x, y), and its value range is The value range of i and j is 0≤x≤M-1, 0≤y≤N-1, and the true phase value can be expressed as:

[0040]

[0041] Where k(x, y) is an integer. Find the pixels in the wrapped phase that have the same phase difference as the unwrapped phase, treat them as a whole phase layer, and add (or subtract) integer multiples of 2π to the whole layer to obtain the phase-recovered unwrapped image. Assuming that the adjustment amount of the nth phase layer is L(n) times 2π, then the elements (x, y) belonging to this layer are:

[0042] k(x, y) = L[n(x, y)] (8)

[0043] Where k(x, y) is the parameter in formula (1). Substituting it into the equation yields:

[0044]

[0045] like Figure 1 As shown, a target pellet interferogram unwrapping method based on edge recognition and phase stratification includes the following steps:

[0046] Step 1: Use the LOG operator to extract edges from the input parcel map. To ensure that all edges are fully identified, a threshold of 0.01 is selected. To remove noise interference, the connected graph is denoised, and connected graphs with less than 500 pixels are removed. Figure 2 This is the edge recognition map of a target pill wrapping map and the connected map before and after denoising.

[0047] Step 2: Select the starting point of each phase layer and ensure that there is only one starting point in the same phase layer.

[0048] The center of the image (i.e., the center of the target pill) is selected as the starting point S(0) of the zero-phase layer through the target pill interference wrapping diagram. A line is drawn outward from this point, which will intersect with the edge of each layer at points P(1), P(2)... and the midpoint of every two adjacent P points is selected as the starting point S(i) of the i-th layer. Figure 3 is the sequence diagram of the selected starting point.

[0049] In practice, phase layers often contain a large amount of fine noise, which can be detected by edge detection operators and mistakenly identified as edges. The solution is to remove these noise using a connected graph algorithm. Morphological closing is then performed on the edge-extracted image to bridge small gaps in the edges and ensure closed edge lines between phase layers.

[0050] The noise in the phase layer is randomly and dispersed, while the real edge between phase layers is a relatively complete circle. Therefore, the connected graph is calculated on the graph after edge recognition, and the connected graph with less than a certain threshold number of points is removed, while the coherent edges are retained.

[0051] Step 3: Identify all phase layers based on the starting point sequence. The specific process is as follows: Figure 4 shown.

[0052] Traverse the selected starting sequence points S(n), perform breadth-first search on each starting point, identify the entire phase layer where it is located, and record the phase layer number of the searched pixel. The specific expression is:

[0053] L[n(x, y)]=i (10)

[0054] Where i is the i-th starting point.

[0055] Step 4: Perform an overall phase shift on the identified phase layer.

[0056] Step 5: Eliminate the remaining transition points, which are caused by the certain width of the edge itself. Figure 5 Phase plots before and after eliminating the residual transition points are shown.

[0057] Compare the phase of all pixels in the entire image with that of the surrounding pixels and compensate for the phase difference. The specific formula is:

[0058]

[0059] Where R is the rounding function, and M is the median phase value within a certain range around the point (x, y).

[0060] Figure 6 The results of applying this method to a target interferogram are shown. (a) and (d) are captured interferograms of the target, (b) and (e) are the corresponding wrapping images and colorbars, and (c) and (f) are the processed results of the unwrapping algorithm and colorbar. The actual processed results show smooth and continuous phase changes, indicating that the method of the present invention has the desired functionality and can quickly and accurately unwrap target interferogram images.

[0061] The embodiments described above provide a detailed description of the technical solutions and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not intended to limit the present invention. Any modifications, supplements and equivalent substitutions made within the scope of the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A target pellet interferogram unwrapping method based on edge recognition and phase stratification, characterized in that: include: (1) Perform edge extraction on the input target pellet interference wrapping pattern to extract the edges between each phase layer; (2) Select the starting point of each phase layer to obtain the starting point sequence S(n) of each phase layer and perform denoising. The specific process is as follows: For the target pellet interference wrapping diagram, the center of the target pellet is selected as the starting point S(0) of the zeroth phase layer. A line is drawn outward from this point, which intersects with the edge of each layer at points P(1), P(2), etc. The midpoint of every two adjacent P points is selected as the starting point S(i) of the i-th layer; thus, the sequence coordinates of the starting points of each phase layer are: Finally, the starting point sequence S(n) of each phase layer is obtained; When performing denoising, a connected graph algorithm is used to remove connected graphs with fewer points than a set threshold, while retaining coherent edges; (3) Traverse the selected starting sequence points S(n), perform breadth-first search on each starting point, identify the entire phase layer where it is located, and record the phase layer number of the searched pixel; (4) Performing an overall phase shift on the identified phase layer to correct the remaining error points in the phase matrix, thereby obtaining the unwrapped phase φ; Perform overall phase shift on the identified phase layer. The specific process is as follows: Record the phase layer number of the searched pixel. The specific expression is: L[n(x,y)]=i Where i is the i-th starting point; and the following formula is used to perform the overall phase shift on the identified phase layer:

2. The target-pill interferogram unwrapping method based on edge recognition and phase stratification according to claim 1, characterized in that: In step (1), the size of the target pellet interference wrapping pattern is M×N, is the wrapping phase of the pixel coordinate (x, y), and its value range is The value range of i,j is 0≤x≤M-1, 0≤y≤N-1.

3. The target-pill interferogram unwrapping method based on edge recognition and phase stratification according to claim 1, characterized in that: In step (1), when performing edge extraction, the LOG operator is used as the edge extraction operator. The LOG operator first performs a Gaussian low-pass filter on the image, which is expressed as: Where f(x,y) is the image and G(x,y) is the Gaussian function, which is expressed as: Where σ is the standard deviation. Since this is a linear system, the order of convolution and differentiation calculation in equation (1) is swapped to obtain: Formula (3) shows that the Gaussian operator is differentiated first and then convolved with the image f(x,y); therefore, the derivative of formula (2) is calculated as: Formula (4) is called the LOG operator.

4. The target-pill interferogram unwrapping method based on edge recognition and phase stratification according to claim 1, characterized in that: In step (4), to correct the remaining error points in the phase matrix, it is necessary to compare the phase of all pixels in the entire image with the surrounding points and compensate for the phase difference. The specific formula is: Where R is the rounding function, and M is the median phase value within a certain range around the point (x, y).

Images