Ghost imaging optimization method based on total intensity sorting of reference light paths

Abstract

The present invention relates to the optimization of ghost imaging detection data, specifically a ghost imaging optimization method based on the total intensity sorting of the reference light path, which uses the statistical characteristics and total intensity of the speckle pattern to sort the original speckle sequence according to the total intensity of the speckle pattern. The new speckle sequence is divided into subsets by a filter to further analyze the effective characteristics of each part of the speckle data. The numerical simulation results show that when the total number of sampling times is 3000 times and the division ratio is 2:1:2, the contrast / mean square error of the proposed method is improved compared with the bucket detector value sorting method, which verifies the effectiveness of the 0.85 sampling rate for speckle screening. At the same time, for the "known" target image, the method in this paper is improved in contrast and mean square error compared with the traditional sorting method at the same ratio. Therefore, the proposed method can be extended to all current association algorithms. In addition, a joint screening method is proposed, which lays a foundation for further research in the future.

Description

Technical Field

[0001] The present invention relates to the screening of ghost imaging speckle patterns, and in particular to a ghost imaging optimization method based on the total intensity sorting of a reference light path. Background Art

[0002] As a novel optical imaging method, ghost imaging has attracted great attention from relevant scholars due to its unique counter-intuitive characteristics that can indirectly reconstruct unknown target objects. Compared with the traditional CCD imaging method, ghost imaging mainly obtains the reconstructed ghost image of the target object by correlating the speckle pattern of the reference light path with the total light intensity value transmitted (or reflected) by the object light path. It has important application value in the fields of nighttime urban security, ultra-low radiation CT medical imaging, etc., which are limited by the array response limit and cannot be directly detected by CCD. In fact, the basic principle of ghost imaging is to use the sequence speckle pattern with different characteristics and the total light intensity value transmitted (or reflected) by the object to realize the inversion of the information of each pixel point of the target object, and the reconstruction effect of the ghost image is better as the number of sequence speckle patterns increases. However, when the reconstruction effect is improved by increasing the number of speckle patterns, it is bound to involve a large amount of measurement and calculation between data, which greatly increases the imaging time and seriously restricts the practical application of ghost imaging.

[0003] In order to reduce the number of measurements and computing time in ghost imaging detection systems, some researchers have proposed in recent years to improve the quality of ghost images by reducing the redundancy between speckle patterns. For example, the target object is illuminated by orthogonalized speckle patterns (such as Hadamard, etc.) to obtain correlated ghost images. This method can achieve complete reconstruction of the target object under full sampling conditions, but as the measurement data decreases, the ghost image will be severely distorted, that is, it is impossible to obtain high-quality ghost images under undersampling conditions. In order to solve the problem of poor ghost image quality under undersampling conditions, the project team proposed a ghost imaging optimization method based on orthogonalized sinusoidal speckle patterns. However, in the actual detection process, due to the limitations or influence of the detection space, environment, and ultra-low radiation modulator, the speckle pattern is often random rather than orthogonal, which makes the above-mentioned ghost imaging quality improvement method based on orthogonal ideas cannot be directly applied to non-orthogonal environments. Therefore, how to reduce the number of random speckle patterns involved in the association and improve the quality of ghost images has become the focus of current research. In the study, it was found that superimposing adjacent speckle patterns can effectively reduce the redundancy between data, and a ghost imaging optimization method based on the combination-modulation of multiple speckle patterns was proposed. This method mainly improves the difference significance between speckle patterns by sorting and combining multiple speckle patterns, thereby reducing the redundancy between speckle patterns, and uses the modulation of bucket detector values ​​to achieve the improvement of ghost image quality.

[0004] In essence, the influence of each speckle pattern on the ghost imaging result is different during the ghost imaging detection process. If the traditional imaging method is used to involve all data in the calculation without distinguishing the influence of these speckle patterns on the imaging effect, it will inevitably inhibit the improvement of the ghost image quality. Therefore, some researchers have proposed to reasonably screen the measurement data before correlation imaging. For example, Wu Lingan et al. used the mean of the bucket detector value as the screening basis, and reduced the influence of large redundant data on the imaging result by deleting the measurement data near the mean, thereby reducing the amount of data involved in the calculation. This method shows that the difference between the speckle patterns after screening based on the mean of the bucket detector value is greater. Theoretically, when the target object is illuminated by the speckle pattern, the average transmittance (or reflectance) of the object can be regarded as an attenuation coefficient, that is, the bucket detector value is the product of the speckle pattern and the average transmittance (or reflectance) of the object. However, the transmittance (or reflectance) of each point of the target object is different, which makes it possible that the bucket detector values ​​are the same or have little difference when the target object is illuminated by the large difference speckle pattern. If the differences between speckle patterns are not distinguished, and data is directly screened based on the bucket detector value, it is bound to cause the deletion of speckle patterns with large differences, which is contrary to the requirement of large-difference speckle patterns in ghost imaging. In addition, since the average transmittance (or reflectance) of the target object is less than 1, the change in the total light intensity value of the bucket detector value is equivalent to the attenuation of the change in the total light intensity value of the speckle pattern. The change in the two values ​​is as follows: Figure 1 As shown in Figure 2, it can be seen that filtering data based on speckle patterns is more advantageous than using bucket detector values. Summary of the invention

[0005] In view of this, this application takes random speckle patterns as a premise, effectively selects speckle patterns by the size of the total light intensity value of the reference light path speckle pattern, and proposes a ghost imaging optimization method based on the total intensity sorting of the reference light path, and uses it in actual engineering practice. This method sorts the speckle pattern in order from small to large according to the total light intensity. The actual analysis and numerical simulation results show that: under different screening ratios, the ghost image effect obtained after filtering the data using the total light intensity value of the speckle pattern is better than the ghost image after filtering the bucket detector value.

[0006] The present invention is implemented by adopting the following technical solution: a ghost imaging optimization method based on the total intensity sorting of the reference light path, comprising the following steps:

[0007] 1) Generate M-beam sequence speckle pattern Here i = 1, 2, ..., M;

[0008] 2) Using the sequential speckle pattern generated in step 1) to illuminate the illuminated object area, using the bucket detector to measure the total intensity value transmitted by the illuminated object area, and obtaining the sequential bucket detector value Here i = 1, 2, ... M;

[0009] 3) Calculate the total light intensity of each sequence speckle pattern in 1) Then the average value of the total light intensity of the speckle pattern is obtained R >, and sort the total intensity values ​​of the M-beam sequence speckle patterns in ascending order, so that the difference between adjacent speckle patterns is reduced;

[0010] 4) Divide the total intensity value of the sorted M-beam sequence speckle pattern into three areas, namely the left area (S Rmin ,S RN ), the area near the mean (S RN ,S RP ) and the right area (S RP ,S Rmax ), S RN ,S RP represents the threshold, S Rmin Indicates the minimum total light intensity after sorting, S Rmax Indicates the maximum total light intensity obtained after sorting.

[0011] 5) Based on the division basis of step 4), the left area (S Rmin ,S RN ) range into subsets {P R1}, the area near the mean (S RN ,S RP ) range into subsets {P R2}, right area (S RP ,S Rmax ) range into subsets {P R3}, and re-divide the corresponding speckle patterns into three subsets: {P R1},{P R2},{P R3}; Then divide the bucket detector values ​​into three subsets according to the three subsets of speckle patterns: {B R1},{B R2},{B R3};

[0012] 6) Using ghost imaging method to analyze the newly acquired speckle pattern subset {P Rj} and the bucket detector value sequence {B Rj}Perform association operation to obtain ghost images. At this time, j=1,2,3.

[0013] The above-mentioned ghost imaging optimization method based on the total intensity ranking of the reference light path is based on the speckle pattern subset {P Rj}, a subset of bucket detector mean screening (traditional screening method) {P Bj ​}Joint screening further reduces data redundancy and improves image quality. The screening formula is: mode-1:I c =P R2 -P R2 ∩P RB2 ,mode-2:I c =(P R1 +P R2 )-P RB2 ,mode-3:I c =(P R1 +P R2 )-P RB2 ∪P R2 , where I c It is the speckle sequence obtained by performing correlation operation on the speckle sequence after the total light intensity of the speckle pattern is screened; P RBj According to P Bj Speckle pattern obtained by sorting; I obtained by selecting one of the screening modes c Perform correlation operation with the corresponding bucket detector value sequence to obtain the ghost image.

[0014] The above-mentioned ghost imaging optimization method based on the total intensity sorting of the reference light path is determined by determining the proportional division parameter To determine the threshold, first set a proportional division parameter Where M represents the total number of speckle patterns, M' represents the number of speckle patterns screened from M, and S is determined based on M'. RN , S RP , partition parameter When the value is 0.8-0.9, the imaging effect is best.

[0015] The principle of the present invention is mainly to use the total light intensity value of the reference light path to select the data with the largest difference between the speckle pattern and the mean of the speckle pattern, so as to improve the imaging quality. That is to say, preprocessing the speckle pattern before reconstructing the target can increase the difference between the associated data. Therefore, the present invention proposes a ghost imaging optimization method based on the total light intensity sorting of the reference light path. The method first sorts the sequence speckle patterns by the total light intensity of the speckle pattern, and uses different association operation rules to achieve the reconstruction of the target object. The numerical simulation results show that for the double-slit binary image, under the total sampling times of 3000 times and the division ratio of 2:1:2, the contrast / mean square error of the proposed method and the method using the bucket detector value sorting are improved by 72.6% / -22.7%, 596.6% / -29.9% and 102.8% / -26% respectively; the speckle near the mean can be deleted at a sampling rate of 0.85, thereby verifying the necessity of effective speckle screening. At the same time, for the "known" target image, under the same sampling times of 3000 times, the division ratio of 2:1:2, and the use of 4 / 10 speckle, the proposed method improves CGI, PNGI, DGI, NGI, PGI, and CS by 33.9 / -8.3%, 31.5% / -7.7%, 13.4% / -1.3%, 14.2% / -1.0%, 56.5% / -15.9%, and 3.8% / -50.1% respectively compared with the traditional CGI, PNGI, DGI, NGI, PGI, and CS. Therefore, the proposed method can be extended to all current association algorithms, and the imaging effect is significantly improved compared with the previous ones. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 Curves for bucket detector intensity fluctuation and speckle pattern total intensity in ascending order.

[0017] Figure 2 Results comparison chart, (a) (b) (c) Original profiles of the two methods, (d) Profile pixel value ascending results of the two methods.

[0018] Figure 3 Schematic diagram of the ghost imaging experimental setup.

[0019] Figure 4 The flowchart of the algorithm for reconstructing the method of the present invention is provided.

[0020] Figure 5 Reconstructed images of different speckle subsets.

[0021] Figure 6 Contrast curves of the reconstructed images at different scales for each reference pattern.

[0022] Figure 7 To compare the reconstructed images using different algorithms.

[0023] Figure 8 CGI simulation imaging of the speckle subsets obtained using Model 1, Model 2, and Model 3.

[0024] Table 1 shows the contrast and mean square error indicators of the traditional sorting method and the proposed method in different subsets.

[0025] Table 2 shows the contrast and mean square error index table of the traditional sorting method and the proposed method under different reconstruction algorithms. DETAILED DESCRIPTION

[0026] Analysis of ghost imaging advantages based on total intensity sorting of reference light paths

[0027] (1) Analysis of the principle of complete reconstruction of ghost imaging

[0028] At present, the main idea of ​​reducing data redundancy is to sort based on the size of the bucket detector value, while the present invention uses the total light intensity of the speckle pattern to sort. In order to better explain the method of the present invention, it is first analyzed here under what circumstances ghost imaging can achieve complete reconstruction.

[0029] Assuming the target is a matrix T(x) of m×n dimensions, and the random measurement matrix φ is of size M×N, where N=m×n, the measurement matrix φ can be obtained:

[0030]

[0031] Here, the measurement matrix φ can be understood as a “large speckle” formed by expanding the speckle pattern into a row and then splicing it together.

[0032] In fact, before the i-th projection, the i-th row of φ should first be rearranged into a two-dimensional pattern to obtain the corresponding Mathematically speaking, if we express is a column vector of size (m×n)×1. After M detections, it must be Has the following linear relationship:

[0033]

[0034] Here, formula (2) can be further expressed as:

[0035]

[0036] Then, extract the first term on the right side of the equal sign of formula (3) and substitute it into formula (1) and formula (2), and we can get:

[0037]

[0038] here, TThe symbol represents the transposition operator. Obviously, φ T The closer φ is to the unit matrix, the higher the quality of the reconstructed image. That is, when φ is an orthogonal matrix, its orthogonal property can help ghost imaging achieve complete reconstruction of the target object.

[0039] (2) Analysis of the advantages of total light intensity sorting of speckle patterns

[0040] Based on the above analysis, the advantages of the total light intensity ranking of the speckle pattern are analyzed by experimental verification. In order to highlight the advantages of the method of the present invention more quickly, it is necessary to determine and Which product of is closer to the unit matrix, so as to judge which method has more advantages. Here, φ is a random matrix of size 3000×4096 generated by simulation. In order to more intuitively show the reconstruction effect of the two methods of total light intensity value sorting and bucket detector value sorting, Figure 2 From one-dimensional curves, two-dimensional matrices and 3D Figure 3 Three results are plotted for each dimension with pixel values ​​normalized to 0-255.

[0041] From other literature, we know that Figure 2 The more prominent the diagonal lines (protruding parts) are and the lower the rest of the background parts are, that is, the closer to the identity matrix, the better the effect. In other words, the diagonal parts play a positive role in image reconstruction, while the non-diagonal areas play a negative role because they generate noise in image reconstruction. At the same time, the higher the noise level, the worse the image quality. Observation Figure 2 It can be found that due to Figure 2 (b) has fewer burrs, so its noise level is lower than 2(a). For easy observation, the present invention extracts the 2000th line and draws the corresponding original profile, as shown in FIG. Figure 2 (c) as shown. Figure 2 In (c), the highest line represents the value on the diagonal, and the disordered area represents the noise level in the non-diagonal area. It can be found that the noise level in the non-diagonal area is very chaotic, and it is not clear which sorting method has a lower noise level. To this end, the present invention arranges the element values ​​in ascending order, and obtains the following Figure 2 (d) The corresponding profile curve is shown in Figure 2. The straight line and dashed line represent (bucket probe value sorting) and The 2000th row element value of (sorted by the present invention). It can be observed that the method proposed by the present invention has a lower noise level, which perfectly explains that the method proposed by the present invention will effectively improve the image reconstruction quality.

[0042] In addition, since the average transmittance (or reflectivity) of the target object is less than 1, the change in the bucket detector value can be regarded as the attenuation of the change in the total light intensity value of the speckle pattern. The trend of the change in the two values ​​is as follows: Figure 1The results show that data screening based on speckle patterns is more advantageous than bucket detector values.

[0043] A method for improving ghost imaging quality based on total intensity sorting of reference light paths

[0044] Specific implementation process of the method

[0045] The schematic diagram of the ghost imaging quality improvement method based on the total intensity sorting of speckles is shown in the figure. Figure 3 As shown in Figure 1, a series of random speckle patterns are irradiated by laser onto the spatial light modulator to simulate the light field in the actual detection environment. Figure 3 The schematic diagram shown can be obtained Figure 4 The flowchart shown in Figure 4 As can be seen from the flow chart, the implementation process of the present invention mainly includes the following six parts, specifically:

[0046] 1) Use a 532nm laser to generate pseudothermal light, and then use DMD to encode the pseudothermal light to generate an M-beam sequence speckle pattern Here i = 1, 2, ... M;

[0047] 2) Using the sequential speckle pattern generated in step 1) to illuminate the illuminated object area, using the bucket detector to measure the total intensity value transmitted by the illuminated object area, and obtaining the sequential bucket detector value Here i = 1, 2, ... M;

[0048] 3) Calculate the total light intensity of each sequence speckle pattern in 1) Then the average value of the total light intensity of the speckle pattern is obtained R >, and sort the total intensity values ​​of the M-beam sequence speckle patterns in ascending order, so that the differences between adjacent speckle patterns are reduced;

[0049] 4) Divide the total intensity value of the sorted M-beam sequence speckle pattern into three areas, namely the left area (S Rmin ,S RN ), the area near the mean (S RN ,S RP ) and the right area (S RP ,S Rmax ), S RN ,S RP represents the threshold, S Rmin Indicates the minimum total light intensity after sorting, S Rmax The present invention is to determine the ratio division parameter To determine the threshold, here we first set a proportional division parameter ​Where M represents the total number of speckle patterns, M' represents the number of speckle patterns screened from M, and S is determined based on M'. RN , S RP .

[0050] 5) Based on the division basis of step 4), the left area (S Rmin ,S RN ) range into subsets {P R1}, the area near the mean (S RN ,S RP ) range into subsets {P R2}, right area (S RP ,S Rmax ) range into subsets {P R3}, and re-divide the corresponding speckle patterns into three subsets: {P R1},{P R2},{P R3}; Then divide the bucket detector values ​​into three subsets according to the three subsets of speckle patterns: {B R1},{B R2},{B R3};

[0051] 6) Using ghost imaging method to analyze the newly acquired speckle pattern subset {P Rj} and the bucket detector value sequence {B Rj}Perform association operation to obtain ghost images. At this time, j=1,2,3.

[0052] Evaluation indicators

[0053] In order to more objectively illustrate the imaging quality of the proposed method, the present invention uses two evaluation indicators, contrast and mean square error, to quantitatively evaluate the reconstructed ghost image, as follows:

[0054] (1) Contrast

[0055] The contrast v in ghost imaging is defined with reference to the Michelson contrast, namely

[0056]

[0057] Among them, G P and G V They are the peak and valley values ​​on the gray value curve of the reconstructed ghost image.

[0058] (2) Mean square error

[0059]

[0060] Among them, G(i,j) is the transmittance of each pixel, and x and y are the sizes of the reconstructed image.

[0061] Numerical simulation and analysis

[0062] (1) Validity Verification

[0063] In order to verify the feasibility and effectiveness of the method proposed in this invention, a binary double-slit image is created as the original image (see Figure 5 (a)) to perform CGI numerical simulation. It should be noted that the experiment is based on different speckle subsets with the same sequence and the same pixel size. Figure 5 The total number of experimental samples is 3000, and the number of speckle measurements for each subset is given below each reconstructed image. Figure 5 (b)-(g) are subsets selected by bucket detector mean (i.e., traditional sorting method), (h)-(m) are subsets selected by the method of the present invention, and (b)-(d) are subsets P B1 , P B2 , P B3 , the division ratio is 2:1:2, (e)-(f) are subsets Q B1 , Q B2 , is the speckle subset after the speckle sequence is divided into 1:1, (g) is the W subset, which contains the speckles in the entire speckle sequence area; (h)-(j) are subsets P R1 , P R2 , P R3 , the division ratio is 2:1:2, (k)-(l) is the subset Q R1 , Q R2 , is the speckle subset after the speckle sequence is divided according to 1:1, (m) is the W subset, which contains the speckles in the entire speckle sequence area;

[0064] observe Figure 5 It can be seen that the quality of the reconstructed images (h)-(m) in the second row is obviously better than that of the reconstructed images (b)-(g) in the first row. Further analysis shows that when the number of measurements is small (N=298 / 301), Figure 5The first row (c) cannot be imaged, while the second row (i) can obtain a clearer ghost image. This experimental result of obtaining clearer images with low sampling data is a major discovery, which can meet the needs of fast and high-quality imaging such as target recognition and tracking. In order to quantitatively analyze the experiment, Table 1 gives the V and MSE values ​​of each reconstructed image. From the overall perspective of Table 1, both the V and MSE indicators have been significantly improved. Compared with the first row, the improvement in the V and MSE values ​​of the reconstructed image in the second row is 72.6% / -22.7%, 596.6% / -29.9%, 102.8% / -26%, 70.7% / -16.0%, 82% / -33.6%, and 0.0% / 0.0%. In terms of the degree of improvement, Figure 5 (m) has no improvement in contrast and mean square error. The main reason is that they use exactly the same speckle set, but with different speckle illumination order. It should be noted that the value of 596.6% in the improvement is large because Figure 5 (c) The image cannot be reconstructed.

[0065] At the same time, by combining Figure 5 From the analysis in Table 1, we can see that using different sorting methods for the same speckle set for reconstruction will also produce significantly different results, which also means that not all speckle data can promote the reconstruction of the target area, which indirectly proves the importance of speckle data screening, that is, effective speckle pattern extraction. Effective speckle pattern screening refers to deleting speckles that contribute less to image reconstruction.

[0066] It is worth mentioning that the ratio used in the above experiment is 2:1:2. But in fact, different division ratios directly affect the quality of the reconstructed image. In order to further screen the effective speckle pattern, here we select ten groups of different ratios for reconstruction while keeping the measurement data the same, and draw their contrast and mean square error curves as shown below: Figure 6 The horizontal axis represents the division ratio of each speckle subset, and the vertical axis represents the contrast (or mean square error) index value. Figure 6 It can be seen that in the reconstruction of each speckle subset, the contrast value and mean square error of the sorting method proposed in the present invention (dark lines) are significantly better than those of the traditional sorting method (light lines). Therefore, it effectively verifies the effectiveness of the sorting method proposed in the present invention in screening speckle patterns. At the same time, we can see that when the division ratio is 0.8-0.9, the contrast value is the largest and the mean square error is the smallest, and the imaging effect is the best at this time. In addition, when the ratio is 0.85, the P value in the method of the present invention is 0.85. R2 The contrast of the subset is compared with the other two subsets P R1 , P R3 Relatively small, mean square error and the other two subsets P R1 , P R3It is relatively large, so it can be deleted, and the remaining speckle patterns are retained.

[0067] (2) Universality analysis

[0068] Next, numerical simulation experiments are carried out to verify the universality of the proposed method. In the experiment, the Chinese character "zhi" of 64 pixel × 64 pixel is selected as the object to be measured (as Figure 7 (a)) for the experiment. By analyzing the speckle subsets obtained under the ratio of 2:1:2, the Figure 7 reconstruction results of different algorithms shown are obtained. Here, the speckles used for reconstruction in the first row and the second row are obtained by screening the bucket detector values and the total intensity of the speckle pattern respectively, and Figure 7 (b)-(g) represent the reconstruction results of CGI, PNGI, DGI, NGI, PGI and CSCGI algorithms respectively. It should be noted that the speckle pattern used in the reconstruction process of the CSCGI method is Figure 7 the set of the first C frames and the last C frames of the speckle sequence obtained in Figure 7 . Therefore, the actual number of measurements is 2×C = 1000. It can be observed

[0069] that the reconstruction image in the first row contains more noise, while the noise in the reconstruction image in the second row is significantly reduced.

[0070] (3) Data joint mining analysis

[0071] In the previous part, the present invention discussed the problem of sorting the total light intensity of speckles. Next, data joint screening is described. First, the and total light intensity sequences of the speckle patterns and the bucket detector value sequences are sorted. Then, three parts of speckle subsets and bucket detector value subsets of the traditional method and the method of the present invention are obtained, and the speckle subsets corresponding to the bucket detector subsets obtained by the traditional method are screened out. Finally, by combining and optimizing the above two speckle sequences, three joint data screening models are introduced. Model 1: Delete the intersection of the P Rj (or P RBj ) subset and the P RBj subset from P Rj ; Model 2: Delete the corresponding P Bj (or P Rj ) subset from P Rj(or P RBj ) subset; Model 3: From P RBj (or P Rj ) to delete P Rj and P RBj The union of .

[0072]

[0073]

[0074]

[0075] Among them, I o and I c are speckle sequences obtained by performing correlation operations on the speckle sequences after screening the bucket detector values ​​and the total light intensity of the speckle pattern; P RBj According to P Bj Sorted speckle patterns; related operations include intersection, difference and union, with symbols of '∩', '-' and '∪', respectively.

[0076] In order to better evaluate the image quality, the present invention uses the CGI algorithm to analyze the 64 pixel × 64 pixel binary double-slit object image (see Figure 8 (a)) is reconstructed. In the case of controlling the variables, the speckle pattern with M = 3000 is obtained as follows Figure 8 The CGI simulation results of the speckle screening subsets of Model 1, Model 2 and Model 3 are shown. Among them, (b)-(d) represent the I obtained using the speckle sequences of Model 1, Model 2 and Model 3. o Reconstruction results, (e)–(g) represent the I obtained using the speckle sequences of model 1, model 2, and model 3. c Reconstruct the result. It should be noted that Figure 8 The numbers below each reconstructed image in the figure indicate the number of speckles. It can be seen that the performance of (e)–(g) is better than that of (b)–(d), and the I c The speckle sequence only needs 482 speckle patterns to obtain good image quality. o The reconstructed image of the speckle sequence does not contain any useful information, which just verifies the hypothesis put forward in the above content.

[0077] Table 1

[0078]

[0079] Table 2

[0080]

Claims

1. A ghost imaging optimization method based on the total intensity ranking of the reference light path, characterized by: The following steps are involved: 1) Generate M-beam sequence speckle pattern Here i = 1, 2, ..., M; 2) Using the sequential speckle pattern generated in step 1) to illuminate the illuminated object area, using the bucket detector to measure the total intensity value transmitted by the illuminated object area, and obtaining the sequential bucket detector value Here i = 1, 2, ..., M; 3) Calculate the total light intensity of each sequence speckle pattern in 1) Then the average value of the total light intensity of the speckle pattern is obtained R >, and sort the total intensity values ​​of the M-beam sequence speckle patterns in ascending order, so that the difference between adjacent speckle patterns is reduced;​ 4) Divide the total intensity value of the sorted M-beam sequence speckle pattern into three areas, namely the left area (S Rmin ,S RN ), the area near the mean (S RN ,S RP ) and the right area (S RP ,S Rmax ), S RN ,S RP represents the threshold, S Rmin Indicates the minimum total light intensity after sorting, S Rmax Indicates the maximum total light intensity obtained after sorting; 5) Based on the division basis of step 4), the left area (S Rmin ,S RN ) range into subsets {P R1 }, the area near the mean (S RN ,S RP ) range into subsets {P R2 }, right area (S RP ,S Rmax ) range into subsets {P R3 }, and re-divide the corresponding speckle patterns into three subsets: {P R1 },{P R2 },{P R3 }; Then divide the bucket detector values ​​into three subsets according to the three subsets of speckle patterns: {B R1 },{B R2 },{B R3 }; 6) Using ghost imaging method to analyze the newly acquired speckle pattern subset {P Rj } and the bucket detector value sequence {B Rj }Perform association operation to obtain ghost images. At this time, j=1,2,3.

2. The ghost imaging optimization method based on the total intensity sorting of the reference light path according to claim 1 is characterized in that: By using the speckle pattern subset {P Rj }, subset of bucket detector mean selection {P Bj }Joint screening further reduces data redundancy and improves imaging quality. The screening formula is: mode-1:I c =P R2 -P R2* P RB2 ,mode-2:I c =(P R1 +P R2 )-P RB2 ,mode-3:I c =(P R1 +P R2 )-P RB2 ∪P R2 , where I c It is the speckle sequence obtained by performing correlation operation on the speckle sequence after the total light intensity of the speckle pattern is screened; P RBj According to P Bj Sorted speckle patterns.

3. The ghost imaging optimization method based on the total intensity sorting of the reference light path according to claim 1 or 2, characterized in that: By determining the proportion of the parameters To determine the threshold, first set a proportional division parameter Where M represents the total number of speckle patterns, M' represents the number of speckle patterns screened from M, and S is determined based on M'. RN , S RP , partition parameter When the value is 0.8-0.9, the imaging effect is best.

Images