Optical imaging system sensor optimization configuration method based on genetic algorithm

Abstract

The present application relates to a genetic algorithm-based optical imaging system sensor optimization configuration method.The present application includes taking the uniformity coefficient of the sensor sensitivity distribution as the optimization target, taking the angle position of the sensor structure light-emitting and light-sensing elements as the optimization object, and obtaining the optimal structure of the optical tomography sensor applicable to the fan-shaped laser through the genetic algorithm.Then, the angles of the light-emitting elements and the light-sensing elements are set according to the optimal sensor structure parameters, so that the optimized sensor is obtained.Finally, the imaging optimization is realized through the imaging system.The present application is not dependent on any specific actual distribution, is applicable to the case without prior information, and has stronger universality.The genetic algorithm is used to realize the optimization of the sensor structure, and the calculation amount of designing an optimal configuration structure with the minimum uniformity coefficient is reduced.

Description

Technical Field

[0001] This invention belongs to the field of optical imaging technology and relates to a method for optimizing the configuration of sensors in an optical imaging system based on a genetic algorithm. Background Technology

[0002] Optical tomography (OPT) based on optical signal attenuation works by obtaining projection information from multiple angles of the cross-section under test and reconstructing the absorption coefficient distribution of the cross-section. As a "hard-field" tomographic technique, OPT boasts advantages such as high frame rate and high spatial resolution, and has the potential for application in challenging industrial fields involving high pressure, high temperature, and complex media. OPT offers advantages such as non-contact measurement, hard field imaging, and high spatiotemporal resolution, demonstrating immense application potential and attracting widespread attention from scholars both domestically and internationally. This technology has already been widely applied in fields such as multiphase flow measurement, internal combustion engine imaging, and functional tissue imaging.

[0003] like Figure 1 As shown, a commonly used optical tomography system based on a ring sensor includes a microcomputer 1, a photodiode array 2, a signal sampling and control unit 3, a power supply 4, and a ring sensor 5. The imaging process is as follows: with the support of the power supply 4, the signal sampling and control unit 3 sends a sampling signal, and the ring sensor 5 begins to operate. After optical signal information is formed on the photodiode array 2, the signal sampling and control unit 3 sends this information to the microcomputer 1 for imaging processing.

[0004] Among them, such as Figure 2 As shown, the ring sensor 5 includes a laser emitting element 6, a photosensitive element 7, and a base 8. During the measurement process, the laser 6 emits a laser beam 9 onto the obstacle, and the photosensitive element 7 absorbs the light signal, forming light signal information on the photodiode array 2. This information is transmitted to the microcomputer 1 by the signal sampling and control unit 3. During the measurement process, the ring sensor 5 can detect light signals at different angles by passing the laser emitting element 6 at different angles, allowing multiple photosensitive elements 7 to detect the light signal at different angles.

[0005] The ring sensor 5 used in the aforementioned optical tomography system contains multiple light-emitting and photosensitive elements. The distribution of these elements determines the sensitivity field distribution of the imaging system, which has a crucial impact on image reconstruction performance. To acquire more projection information and optimize image quality, the traditional optimization method is to increase the number of light-emitting and photosensitive elements in the sensor. However, on the one hand, in applications where the number of measurement paths in the area to be measured is limited, simply increasing the number of light-emitting and photosensitive elements cannot improve image reconstruction performance; on the other hand, when the number of light-emitting and photosensitive elements is too large, the overall system energy consumption is too high, and too much data is collected, which may affect the operation of the entire system.

[0006] Currently, research on the impact of changing the arrangement of sensor light-emitting and photosensitive elements on imaging performance is insufficient. Therefore, combining the aforementioned imaging system, it is feasible and meaningful to optimize sensor imaging quality by changing the placement angle and position of these two types of elements while keeping the number of light-emitting and photosensitive elements constant. Summary of the Invention

[0007] To overcome the shortcomings of existing technologies, this invention proposes a sensor structure optimization method for optical tomography systems based on a genetic algorithm. First, the uniformity coefficient of the sensor sensitivity distribution is used as the optimization objective, and the angular positions of the emitting and photosensitive elements in the sensor structure are used as optimization objects. A genetic algorithm is then used to obtain an optimized optical tomography sensor structure suitable for fan-shaped lasers. Next, the angles of the emitting and photosensitive elements are set according to the optimal sensor structure parameters, thus obtaining the optimized sensor. Finally, imaging optimization is achieved through the imaging system.

[0008] The present invention specifically includes the following steps:

[0009] Step 1: Define the uniformity coefficient Used to evaluate the uniformity of sensitivity distribution. Among them, S is the average value of the pixel sensitivity distribution. d denoted as the standard deviation of the sensitivity distribution.

[0010] Step 2: Define the sensor structure parameter vector Where θ1,...,θ n These represent the angular positions of the n laser light-emitting elements. These represent the angular positions of the m photosensitive elements.

[0011] Step 3: Using the uniformity coefficient S u As a reference metric, the sensor structure parameter vector R is optimized. A genetic algorithm is used to obtain an optimal configuration structure R with the minimum uniformity coefficient. bt .

[0012] Specifically:

[0013] Step 1: Randomly generate p initial structure vectors Each initial vector is encoded as a chromosome, and the iteration counter is set to t=0. This represents the p-th chromosome in the t-th generation structure vector;

[0014] Step 2: Calculate the initial structure vector for each generation t. Individual evenness parameters The sum of uniformity parameters S of all individuals sum Calculate each structure vector relative uniformity parameters Based on the proportional selection operator, the two chromosomes selected are... i∈[1,p], j∈[1,p], i≠j, and then the single-point crossover operator and mutation operator perform the following operations on these two chromosomes to produce offspring:

[0015] First, the single-point crossover operator uses a crossover probability α to allow gene exchange at random crossover points on two parent chromosomes.

[0016] Then, the mutation operator inverts a gene at a random position on the chromosome after gene exchange with a mutation probability β, resulting in a new chromosome;

[0017] Step 3: Repeat the selection, crossover, and mutation operations until new offspring are reconstructed;

[0018] Step 4: Replace the parent chromosome with the newly formed offspring chromosome, and increment the iteration count by 1;

[0019] Step 5: If the sum of the uniformity parameters of the resulting offspring is less than the expected threshold or the maximum number of iterations is reached, exit and output R. bt Otherwise, proceed to step 2;

[0020] Step 4: Finally, based on the optimal configuration R bt By setting the sensor's light emission and the angle distribution of the photosensitive elements, the structural configuration of the imaging system can be optimized.

[0021] The advantages of this invention compared to the prior art are as follows:

[0022] (1) When the number of measurement paths in some areas to be measured is limited, the optimized imaging system can acquire more projection information without increasing the number of light-emitting elements and photosensitive elements, thereby improving the reconstruction quality of the image;

[0023] (2) The optical tomography structure optimization method based on genetic algorithm proposed in this invention does not depend on any specific actual distribution, is applicable to situations without prior information, and has stronger universality.

[0024] (3) The optimized sensor beams of this invention have a more uniform arrangement and distribution, resulting in images with better reconstruction accuracy and lower noise. Compared with regular and random sensor configurations, the optimized structure can more clearly reconstruct geometric structures, hard boundaries and blurred boundaries with different distributions.

[0025] (4) Due to the infinite combinations of sensor angles, designing an optimal configuration structure with the minimum uniformity coefficient requires a lot of computation. The genetic algorithm used in this invention optimizes the sensor structure and reduces the amount of computation required to design an optimal configuration structure with the minimum uniformity coefficient. Attached Figure Description

[0026] Figure 1 An optical tomography system based on a ring sensor;

[0027] Figure 2 This is a schematic diagram of a ring-shaped optical sensor structure.

[0028] Figure 3 To find the optimal configuration structure R bt A flowchart;

[0029] Figure 4 A structural comparison diagram of the sensor and the optimized sensor;

[0030] Figure 5 Comparison of Landweber algorithm imaging before and after sensor optimization. Detailed Implementation

[0031] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. The description of the embodiments of the present invention is only a preferred embodiment of the present invention and should not be regarded as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should fall within the scope of the present invention.

[0032] For example Figure 2 The ring sensor shown illustrates a sensor optimization configuration method for an optical imaging system based on a genetic algorithm. The specific steps are as follows:

[0033] Step 1: Define the uniformity coefficient Uniformity coefficient S u It is widely used to evaluate the uniformity of sensitivity distribution. Among them, S is the average value of the pixel sensitivity distribution. d Let be the standard deviation of the sensitivity distribution. The sensitivity distribution has a significant impact on image reconstruction quality. A uniform sensitivity distribution can effectively improve the quality of the reconstructed image and avoid distortion; while a non-uniform sensitivity distribution will worsen the ill-conditioned nature of the inverse problem and reduce the reconstruction quality.

[0034] Uniformity coefficient S of sensitivity distribution u The uniformity coefficient S is directly related to the accuracy of the reconstructed image. u The sensor structure has a lower uniformity coefficient S, which can better characterize the reconstruction error. u The higher the homogeneity coefficient S, the higher the quality of the reconstructed image. u As a functional predictor of image reconstruction performance.

[0035] Step 2: Define the sensor structure parameter vector Where θ1,...,θ n These represent the angular positions of the n laser light-emitting elements. Let R represent the angular positions of the m photosensitive elements. Since both the light-emitting and photosensitive elements are arranged on the same ring-shaped sensor, i.e., on the same arc, a unified starting 0-degree position is sufficient to determine the position of an element using a single radian. If the divergence angle of the light-emitting elements is disregarded, and only the arrangement of the light-emitting and photosensitive elements is considered, their angular information is sufficient to represent all the structural information of the sensor. Therefore, the sensor structural parameter vector R is the object of optimization.

[0036] Step 3: Using the uniformity coefficient S u As a reference indicator, the sensor structure parameter R is optimized. A genetic algorithm is used to obtain an optimal configuration structure R with the minimum uniformity coefficient. bt In this embodiment, the number of light-emitting elements and photosensitive elements is set to m, the divergence angle is γ, the maximum number of iterations is num, the initial number of structures is p (default value), the crossover probability is α, and the mutation probability is β. To balance the search area's capability and profitability, α ∈ [0.25, 0.9] and β ∈ [0.01, 0.2]. Simultaneously, an expected threshold S is set. umin As a criterion for judging the best solution.

[0037] like Figure 3 As shown, the optimal configuration structure R is sought. bt The specific process is as follows:

[0038] 1. Randomly generate p initial structure vectors Each structure vector is encoded as a chromosome, and an iteration counter is set to t=0. This represents the p-th chromosome in the t-th generation structure vector;

[0039] Step 2: Calculate the initial structure vector for each generation t. Individual evenness parameters The sum of uniformity parameters S of all individuals sum Calculate each structure vector relative uniformity parameters The proportional selection operator is the ratio of the probability of an individual being selected and passed on to the next generation of the population to the relative evenness coefficient of that individual. Based on the principle that the size is inversely proportional, two chromosomes were selected. i∈[1,p], j∈[1,p], i≠j, and then the single-point crossover operator and mutation operator perform the following operations on these two chromosomes to produce offspring:

[0040] First, the single-point crossover operator uses a crossover probability α to allow gene exchange at random crossover points on two parent chromosomes.

[0041] Then, the mutation operator inverts a gene at a random position on the chromosome after gene exchange with a mutation probability β, resulting in a new chromosome;

[0042] 3. Repeat selection, crossover, and mutation operations until new offspring are reconstructed;

[0043] 4. Replace the parent chromosome with the newly formed offspring chromosome, and increment the iteration count by 1;

[0044] 5. If a globally optimal solution is found, that is, the sum of the uniformity parameters S of the resulting offspring... sum Less than the expected threshold S umin If the maximum number of iterations num is reached, the loop exits, and the optimized structure vector at this point is returned. As R bt Output the output; otherwise, proceed to step 2.

[0045] Step 4: Finally, based on the optimal configuration R bt By adjusting the sensor's light emission and the angular distribution of the photosensitive elements, the structure of the imaging system can be optimized. For example... Figure 2 As shown, the laser light-emitting element 6 and the photosensitive element 7 are positioned at variable angles, according to... By obtaining the one-to-one correspondence between the optimal vector and the placement angles of the light-emitting and photosensitive elements, and setting the angles sequentially, the sensor with optimized structure can be obtained.

[0046] like Figure 4 As shown, the unoptimized sensor 10 is a regular sensor, meaning that the photosensitive element 11 and the light-emitting element 12 are alternately and equally spaced. The reconstructed image obtained using the light information from this sensor has poor reconstruction accuracy and high noise. The optimized sensor 13 is an irregular sensor, meaning that the photosensitive element 14 and the light-emitting element 15 are not equally spaced, but rather arranged according to the optimal configuration R. bt Place it according to the provided location information.

[0047] The optimal sensor parameter configuration is calculated by computer 1, and the ring sensor 5 is adjusted to achieve structural optimization, after which it can be used based on... Figure 1 The optical tomography experimental system shown is used for imaging, and the specific steps are as follows:

[0048] Step 1. Based on the optimal configuration R obtained by the genetic algorithm btThe sensor element distribution is set, and the optimized imaging system is used for imaging operations. First, the sensitivity field matrix A is calculated based on the sensor size and the positions of the light-emitting and photosensitive elements. The light emission information detected by the photosensitive element can be represented as b = Ax, where b and x are both column vectors, where the elements in b represent the projections in each direction, and x is the original distribution of the absorption coefficient, A ∈ R. J×I This represents the sensitivity matrix, where element l ji Let represent the chord length of the j-th ray within the i-th pixel. A two-dimensional absorption coefficient distribution image is reconstructed using a large amount of optical projection information b and a sensitivity matrix A. Tomographic imaging is achieved by solving the inverse problem of b = Ax.

[0049] Step 2. Light up n lasers sequentially, measure the signals of m photosensitive elements respectively, and form a background signal S. b Without placing any physical object, lasers are sequentially illuminated along a certain direction. At any given time, only one of the n lasers can be on. Once one laser is on, each of the m photosensitive elements can collect a signal data point, resulting in n×m data points per cycle. The collected environmental projection data without any physical object is used to form a background signal S. b .

[0050] Step 3. Place targets of different shapes within the cross section to be measured to form different cross section distributions.

[0051] Step 4. Light up the lasers sequentially and measure the measurement signal S generated by the photosensitive element. m After placing the object, repeat step 2, and use the collected environmental projection data at the time of object placement to form a measurement signal S. m .

[0052] Step 5. Calculate and obtain the projection signal P = ln(S) b / S m The original measurement signal is filtered according to the formula P = ln(S). b / S m This yields a (n×m)×1 projection data matrix column vector P.

[0053] Step 6. Reconstruct the cross-sectional image using the Landweber algorithm. In this paper, the initial value selected for this algorithm is the LBP algorithm reconstruction result: x = A T P, where P is the column vector of the projected data matrix, and A is the sensitive field matrix under this structure. The LBP algorithm essentially uses A... T As A -1 Solve the image grayscale equation using an approximate matrix.

[0054] The basic formula for the Landweber algorithm is: X k+1 =Xk -ηA T (AX k -P), where η represents the step size and k represents the number of iterations. The Landweber algorithm solves the problem by finding the optimal solution step by step within the solution space. The step size controls the scope of the search during the iteration process. When the step size is too large, it is easy to miss the optimal solution during the search; when the step size is too small, it wastes search time and reduces efficiency. The termination condition of the algorithm is generally selected as the difference between the solutions of two adjacent iterations, i.e., ||X|. k+1 -X k ||<δ, where the value of δ is very small, and the termination condition is determined by the actual situation.

[0055] If the algorithm reaches the stopping condition after iteration, it exits the iteration; otherwise, it continues running. The grayscale vector P obtained at the end of the iteration can be used to reconstruct the cross-sectional image of the pipe.

[0056] To verify the effectiveness and superiority of the optimized sensor structure configuration, an optimized sensor was obtained by optimizing a conventional sensor according to the implementation steps, and then the Landweber algorithm was used to reconstruct the images.

[0057] like Figure 5 As shown, Figure 5 In the image, (a)-(e) represent five absorption coefficient distributions, serving as reference images for measuring the quality of the reconstructed image. (a)-(e) represent a single phantom with "hard edges", a double phantom with "hard edges", a Gaussian phantom with "soft edges", a cross distribution with "hard edges", and a concave distribution with "hard edges", respectively. Figure 5 In the figure, (f)-(j) and (k)-(o) correspond to the image reconstruction results based on the sensor structure before and after optimization, respectively. As shown in the figure, the image based on the optimized sensor structure can more clearly and accurately describe the geometric image and edge information, while the reconstruction error is lower than that of the image based on the sensor structure before optimization. Even with complex distributions, the optimized structure can still accurately reconstruct the distribution state, with clear boundaries and less noise.

[0058] Therefore, the sensor configuration optimization method based on genetic algorithms can effectively improve the image reconstruction quality.

Claims

1. A sensor optimization configuration method for an optical imaging system based on a genetic algorithm, characterized in that: Step 1: Define the uniformity coefficient Used to evaluate the uniformity of sensitivity distribution; among which, S is the average value of the pixel sensitivity distribution. d The standard deviation of the sensitivity distribution; Step 2: Define the sensor structure parameter vector Where θ1,...,θ n These represent the angular positions of the n laser light-emitting elements. These represent the angular positions of the m photosensitive elements; Step 3: Using the uniformity coefficient S u As a reference indicator, the sensor structural parameter R is optimized; a genetic algorithm is used to obtain an optimal configuration structure R with the minimum uniformity coefficient. bt ; Specifically: Step 1: Randomly generate p initial structure vectors Each initial vector is encoded as a chromosome, and the iteration counter is set to t=0. This represents the p-th chromosome in the t-th generation structure vector; Step 2: Calculate the initial structure vector for each generation t. Individual evenness parameters The sum of uniformity parameters S of all individuals sum Calculate each structure vector relative uniformity parameters Based on the proportional selection operator, the two chromosomes selected are... Then, the single-point crossover operator and the mutation operator are used to perform the following operations on the two chromosomes to produce offspring: First, the single-point crossover operator uses a crossover probability α to allow gene exchange at random crossover points on two parent chromosomes. Then, the mutation operator inverts a gene at a random position on the chromosome after gene exchange with a mutation probability β, resulting in a new chromosome; Step 3: Repeat the selection, crossover, and mutation operations until new offspring are reconstructed; Step 4: Replace the parent chromosome with the newly formed offspring chromosome, and increment the iteration count by 1; Step 5: If the sum of the uniformity parameters of the resulting offspring is less than the expected threshold or the maximum number of iterations is reached, exit and output R. bt Otherwise, proceed to step 2; Step 4: Based on the optimal configuration R bt By setting the sensor's light emission and the angle distribution of the photosensitive elements, the structural configuration of the imaging system can be optimized.

Images