A tomographic imaging method to overcome internal light path occlusion

Abstract

The present invention provides a tomography method for overcoming internal light path occlusion, discretizing a three-dimensional signal field to be measured, and then classifying and processing the voxels obtained by discretization and the pixels on the two-dimensional projection, pre-excluding a part of invalid voxels with a signal value of 0 and zero-value pixels on the two-dimensional projection, and reconstructing the light field distribution of the three-dimensional signal field only according to the mapping relationship between valid voxels and non-zero-value pixels, thereby greatly improving the calculation efficiency and reconstruction accuracy of the three-dimensional tomography, and being able to greatly reduce the reconstruction errors caused by incomplete projections; in addition, the present invention can be applied not only to tomography when there are occluders inside the three-dimensional signal field to be measured, but also to tomography when there are occluders in the open light path during the light propagation process.

Description

Technical Field

[0001] The invention belongs to the technical field of optical imaging, and in particular relates to a tomographic imaging method that overcomes internal light path shielding. Background Art

[0002] Three-dimensional tomography technology reconstructs the three-dimensional spatial distribution of the target to be measured based on projections obtained from simultaneous measurements of different perspectives. It is widely used in flow and combustion diagnosis problems due to its non-contact nature and high spatiotemporal resolution. According to the type of measurement environment, tomography problems can be divided into two types: confined space and open space. The signal emitted by the target to be measured in the open space is directly received by the image sensor, which is often used to study the three-dimensional distribution of flame combustion and fluid flow characteristics, such as velocity field, concentration, flame surface characteristics and temperature. However, in actual measurements, combustion and flow often occur in confined spaces, and the environmental conditions, flow and flame presentation patterns are more complex than those in open spaces, which brings great challenges to the analysis of flow or combustion mechanisms. However, in the previous development of three-dimensional tomography algorithms, people have not discussed much about the impact of internal optical path occlusion (IOB) on actual devices in the optical field. The IOB can be any opaque mechanical part of the device under study, such as the shaft in tunnel flow measurement, the turbine rotor guide vane in turbine flow field measurement, etc. It will prevent part of the light source in the 3D light field from being transmitted to the optical sensor, thereby affecting the integrity of the information in the 2D projection collected by the image sensor, and further leading to the wrong reconstruction of the 3D signal. For example, when performing tomographic measurement of the Mie scattering signal distribution generated by the injection of fuel spray in a gasoline optical engine, the spark plug in the combustion chamber becomes the IOB in the measurement process. Therefore, no matter what sensor orientation is chosen, the projection recorded through the engine optical cylinder will contain incomplete information. Summary of the invention

[0003] In order to solve the problem of three-dimensional tomography with internal occlusion, the present invention provides a tomography method that overcomes internal light path occlusion, which can greatly improve the calculation speed and reconstruction accuracy of three-dimensional tomography.

[0004] A tomographic imaging method for overcoming internal light path occlusion comprises the following steps:

[0005] Discretizing a three-dimensional signal field including a target to be measured and an obstruction into a plurality of voxels to obtain a discrete signal field to be measured, wherein the target to be measured is a transparent target or a semi-transparent target;

[0006] Obtaining the spatial position coordinates of the obstruction in the discrete signal field to be tested;

[0007] Three or more image sensors arranged around the three-dimensional signal field simultaneously acquire a two-dimensional projection of the three-dimensional signal field;

[0008] Valid voxels are screened out according to the distribution of non-zero value pixels on the two-dimensional projection and the spatial position coordinates of the occluder, wherein the valid voxels satisfy: when performing ray tracing on the valid voxels, there is at least one image sensor such that the light path between the valid voxels and the non-zero value pixels on the two-dimensional projection acquired by the image sensor is not blocked by the occluder;

[0009] The light field distribution of the three-dimensional signal field is reconstructed according to the mapping relationship between the effective voxels and the non-zero value pixels on the two-dimensional projection of each image sensor, so as to achieve the tomographic imaging of the target to be measured.

[0010] Furthermore, the reconstruction method of the light field distribution of the three-dimensional signal field is as follows:

[0011] Construct the following iterative equation:

[0012]

[0013] ΔP(k) I =P(k) I -P 实 (k)

[0014]

[0015] Among them, F(q) I+1 is the signal intensity of the qth valid voxel in the I+1th iteration, F(q) I is the signal intensity of the qth valid voxel in the Ith iteration, q = 1, 2, …, Q′, Q′ is the total number of valid voxels, α is the set relaxation factor, K′ is the total number of non-zero pixels contained in all two-dimensional projections, PSF is the mapping relationship matrix between each non-zero pixel and each valid voxel, ‖·‖ is the modulus operation, PSF(k,q) is the mapping value between the kth non-zero pixel and the qth valid voxel, ΔP(k) I is the simulated intensity P(k) obtained by the kth non-zero pixel in the Ith iteration I Compared with the actual strength P measured experimentally 实 (k) The error between

[0016] Set the initial value F(q) for the signal intensity of each valid voxel 0 Then, the iteration equation is continuously iterated until the difference between the signal intensity of each effective voxel obtained by the iteration and the set convergence value is less than the set value, thereby completing the reconstruction of the light field distribution of the three-dimensional signal field.

[0017] Furthermore, the method for obtaining the spatial position coordinates of the obstruction in the discrete signal field to be measured is:

[0018] If the design drawings of the three-dimensional signal field are unknown, multiple projections of the occluder in different directions are obtained by directly photographing the occluder, and then the occluder is reconstructed based on the projections to obtain its spatial position coordinates in the discrete signal field to be tested; if the design drawings of the three-dimensional signal field are known, the spatial position coordinates of the occluder in the discrete signal field to be tested are obtained by checking the design drawings.

[0019] Furthermore, after the spatial position and shooting direction of each image sensor are calibrated, the calibrated image sensors are used to simultaneously acquire the two-dimensional projection of the three-dimensional signal field.

[0020] Furthermore, the image sensor is a camera or a fiber bundle endoscope.

[0021] Beneficial effects:

[0022] 1. The present invention provides a tomographic imaging method for overcoming internal light path occlusion, discretizing a three-dimensional signal field to be measured, and then classifying and processing the voxels obtained by discretization and the pixels on the two-dimensional projection, pre-excluding a part of invalid voxels with a signal value of 0 and zero-value pixels on the two-dimensional projection, and reconstructing the light field distribution of the three-dimensional signal field only according to the mapping relationship between valid voxels and non-zero-value pixels, thereby greatly improving the calculation efficiency and reconstruction accuracy of three-dimensional tomographic imaging, and can greatly reduce the reconstruction errors caused by incomplete projections; in addition, the present invention can be applied not only to tomographic imaging when there are occluders inside the three-dimensional signal field to be measured, but also to tomographic imaging when there are occluders in the open light path during the light propagation process.

[0023] 2. The present invention provides a tomographic imaging method that overcomes internal light path occlusion. When calculating the signal value of a valid voxel through an iterative method, only non-zero-value pixels and valid voxels are used to construct a mapping relationship matrix between the two, which significantly reduces the number of voxels to be iteratively solved and the dimension of the mapping relationship matrix, thereby greatly improving the efficiency and accuracy of three-dimensional tomographic calculations.

[0024] 3. The present invention provides a tomographic imaging method that overcomes internal light path obstruction. After calibrating the spatial position and shooting direction of each image sensor, the calibrated image sensor is used to simultaneously obtain the two-dimensional projection of the three-dimensional signal field, which can further improve the reconstruction accuracy of the tomographic imaging. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flowchart of the IOB algorithm provided by the present invention;

[0026] Figure 2 A schematic diagram of ray tracing of the IOB algorithm provided by the present invention. DETAILED DESCRIPTION

[0027] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0028] The present invention proposes a new three-dimensional tomography method, which is mainly used for three-dimensional tomography problems when there is internal optical blockage IOB (Internal Optical Blockage). The method is based on the principle of rectilinear propagation of light. First, according to the spatial position relationship between the image sensor and the internal blockage of the three-dimensional signal field to be measured, the occlusion relationship of the voxels in the three-dimensional signal field to be measured is judged and the mapping relationship matrix PSF of the imaging system is corrected. Then, according to the two-dimensional projections obtained from different azimuths, the voxels are divided into two categories: one is the voxels whose signal value can be directly determined to be 0, without participating in the subsequent iterative solution process; the other is the valid voxels, whose signal value size cannot be obtained in advance and can only be solved through the subsequent iterative process. Finally, according to the mapping relationship matrix PSF between the pixels whose signals are not 0 and the valid voxels, the three-dimensional signal distribution of the target to be measured can be obtained by iterative solution.

[0029] Specifically, Figure 1 As shown, a tomographic imaging method for overcoming internal light path occlusion comprises the following steps:

[0030] S1: discretizing a three-dimensional signal field including a target to be measured and an obstruction into a plurality of voxels to obtain a discrete signal field to be measured, wherein the target to be measured is a transparent target or a semi-transparent target.

[0031] It should be noted that the three-dimensional signal field is discretized into multiple voxels, that is, the three-dimensional signal field is gridded to obtain multiple discrete volume elements. At the same time, the present invention does not make specific restrictions on the shape of the volume element, and it can be rectangular, square, triangular or polygonal, as long as the discretization of the three-dimensional signal field is achieved.

[0032] S2: Obtain the spatial position coordinates of the occluder in the discrete signal field to be tested.

[0033] It should be noted that the specific position of the obstruction in the discrete signal field to be tested can be determined in the following two ways. In the case of a design drawing of an unknown three-dimensional signal field, multiple projections of the obstruction in different directions can be obtained by directly taking pictures of the obstruction, and then its specific shape and coordinates can be obtained by reconstruction. In the case of known equipment drawings, its specific shape and size can also be obtained by viewing the data drawings. After obtaining the shape and size of the obstruction, the obstruction can be discretized to obtain its position coordinate distribution in the three-dimensional signal field to be tested, in preparation for subsequent judgment of the obstruction conditions.

[0034] S3: Three or more image sensors arranged around the three-dimensional signal field simultaneously acquire a two-dimensional projection of the three-dimensional signal field.

[0035] It should be noted that in order to further improve the reconstruction accuracy of tomographic imaging, the present invention can calibrate the spatial position and shooting direction of each image sensor, and then use the calibrated image sensor to simultaneously obtain the two-dimensional projection of the three-dimensional signal field. Among them, the present invention is the same as the calibration method of the traditional tomographic algorithm. Generally, a checkerboard acrylic plate with a known pattern and actual size is placed in the center of the three-dimensional signal field to be measured as a calibration target, and then multiple image sensors arranged around it are used to take pictures of it to obtain its two-dimensional projection in different orientations. Since the actual size of the checkerboard is known, the distance and relative azimuth relationship between the image sensor and the target object (acrylic checkerboard plate) can be calculated by comparing the difference between the pattern size and the actual size in the image.

[0036] After the calibration is completed, the calibration target (i.e., the acrylic checkerboard) is removed, and then all the image sensors arranged around the three-dimensional signal field are used to simultaneously take pictures of the three-dimensional signal field with the obstruction, and the transient projections of the three-dimensional signal field to be measured in different orientations are obtained.

[0037] Optionally, the image sensor used in the present invention is a camera or a fiber optic bundle endoscope.

[0038] S4: Filter out valid voxels according to the distribution of non-zero value pixels on the two-dimensional projection and the spatial position coordinates of the occluder, wherein the valid voxels satisfy the following conditions: when ray tracing is performed on the valid voxels, there is at least one image sensor such that the light path between the valid voxels and the non-zero value pixels on the two-dimensional projection acquired by the image sensor is not blocked by the occluder.

[0039] For example, Figure 2As shown in the figure, a rectangular measurement signal field contains two signal models (i.e., the square in the lower right corner and the triangle area in the upper left corner) and an occluder (i.e., the black line segment in the middle), all of which are discretized into small rectangular voxels (not drawn for the simplicity of the figure). Three cameras are installed around the measurement area, and based on the principle of parallel beam imaging, the two-dimensional projection of the model is obtained on the camera CCD array. The pixels on each CCD array can be divided into two categories: non-zero value pixels that can receive light signals and pixels with zero signal values ​​that cannot receive signals, such as the pixel division on the CCD array of camera No. 2. Based on this division, the ray tracing of edge pixels (such as l1 and l2 on camera 2) is drawn with gray dotted lines. Guided by these lines, valid voxels can be accurately distinguished from the measurement domain according to the following criteria: when ray tracing is performed on any voxel, the light of the voxel can directly reach at least a non-zero value pixel area of ​​a camera that can provide "direct observation", and the voxel can be judged as a valid voxel. It should be noted that "direct observation" means that the ray tracing from a specific voxel will not be blocked by an occluder. According to this criterion, the valid voxels in the measurement domain are marked in the shaded area. At the same time, all voxels in the non-shaded area correspond to the pixel area with zero signal value on at least one camera that can be directly observed, so that their signal value can be determined to be zero and excluded from the subsequent iterative calculation.

[0040] It should be noted that when establishing a ray tracing relationship to screen valid voxels, it is also possible to start from the non-zero value pixels on the two-dimensional projection, which is not described in detail in the present invention. In addition, the present invention can also obtain the number of direct observation cameras n for each valid voxel. d Then, according to the number of direct observation cameras n d To adjust the arrangement of cameras so that voxels can be observed by more cameras, thus achieving better reconstruction effect.

[0041] S5: Reconstruct the light field distribution of the three-dimensional signal field according to the mapping relationship between the effective voxels and the non-zero value pixels on the two-dimensional projection of each image sensor to achieve tomographic imaging of the target to be measured; wherein the reconstruction method of the light field distribution of the three-dimensional signal field is as follows:

[0042] S51: Construct the following iterative equation:

[0043]

[0044] ΔP(k) I =P(k) I -P 实 (k)

[0045]

[0046] Where I represents the number of iterations, F(q) I+1is the signal intensity of the qth valid voxel in the I+1th iteration, F(q) I is the signal intensity of the qth valid voxel in the Ith iteration, q = 1, 2, …, Q′, Q′ is the total number of valid voxels, α is the relaxation factor that determines the convergence speed of the iteration, usually ranging from 0 to 2, K′ is the total number of non-zero pixels contained in all two-dimensional projections, PSF is the mapping relationship matrix between each non-zero pixel and each valid voxel, ‖·‖ is the modulus operation, PSF(k,q) is the mapping value between the kth non-zero pixel and the qth valid voxel, ΔP(k) I is the simulated intensity P(k) obtained by the kth non-zero pixel in the Ith iteration I Compared with the actual strength P measured experimentally 实 (k) The error between

[0047] S52: Set the initial value F(q) for the signal strength of each valid voxel 0 Then, the iteration equation is continuously iterated until the difference between the signal intensity of each effective voxel obtained by the iteration and the set convergence value is less than the set value, thereby completing the reconstruction of the light field distribution of the three-dimensional signal field.

[0048] That is to say, after step S4, the position of the effective voxel has been determined. By correcting the traditional mapping relationship matrix between voxels and pixels, the mapping relationship matrix is ​​modified to only consider the mapping relationship between effective voxels and non-zero value pixels, so that it can be combined with the occlusion judgment condition; after obtaining the corresponding relationship between the effective voxels and the non-zero value pixels that are not 0, the signal distribution of the three-dimensional signal field can be solved by an iterative method, and after several iterations, the signals of all effective voxels will converge to a stable value, which is the signal value of the effective voxel, that is, the reconstruction result of the three-dimensional signal field to be measured, and finally the reconstruction of the light field is completed.

[0049] It can be seen that the present invention proposes an IOB imaging algorithm based on the traditional open space tomography algorithm, which can correctly reconstruct the three-dimensional signal field according to the incomplete two-dimensional projection of the signal. The first step of the IOB imaging algorithm is to establish a ray tracing relationship, starting from the pixel on the projection, through the center of the lens to the unique correspondence of the discrete voxels in the signal field to be measured, so as to determine whether the light path will be blocked by internal obstructions. The second step is to divide the voxels into two categories according to the two-dimensional projections obtained from different azimuths: one is the invalid voxels whose signal value can be directly determined to be 0, without participating in the subsequent iterative solution process; the other is the valid voxels, whose signal value size cannot be obtained in advance and can only be solved through the subsequent iterative process. The third step is to iteratively solve according to the mapping relationship between the non-zero value pixels whose signals are not 0 on the two-dimensional projection and the valid voxels, and the three-dimensional signal distribution of the target to be measured can be obtained.

[0050] Compared with the reconstruction formed by the traditional tomography algorithm, the IOB algorithm of the present invention corrects the mapping relationship between the three-dimensional signal field and its two-dimensional projection when there is an obstruction, which can solve the reconstruction error caused by incomplete projection; at the same time, the voxels of the three-dimensional signal field to be measured are classified, and a part of the invalid voxels with a determined signal value of 0 are pre-excluded, which can greatly improve the calculation speed and reconstruction accuracy. In addition, the present invention is not limited to obstructions inside the measured domain, but is also applicable to obstructions in the open optical path propagation process.

[0051] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art may certainly make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. A tomographic imaging method for overcoming internal light path occlusion, characterized in that: The following steps are involved: Discretizing a three-dimensional signal field including a target to be measured and an obstruction into a plurality of voxels to obtain a discrete signal field to be measured, wherein the target to be measured is a transparent target or a semi-transparent target; Obtaining the spatial position coordinates of the obstruction in the discrete signal field to be tested; Three or more image sensors arranged around the three-dimensional signal field simultaneously acquire a two-dimensional projection of the three-dimensional signal field; Valid voxels are screened out according to the distribution of non-zero value pixels on the two-dimensional projection and the spatial position coordinates of the occluder, wherein the valid voxels satisfy: when performing ray tracing on the valid voxels, there is at least one image sensor such that the light path between the valid voxels and the non-zero value pixels on the two-dimensional projection acquired by the image sensor is not blocked by the occluder; The light field distribution of the three-dimensional signal field is reconstructed according to the mapping relationship between the effective voxels and the non-zero value pixels on the two-dimensional projection of each image sensor, so as to realize the tomographic imaging of the target to be measured; The reconstruction method of the light field distribution of the three-dimensional signal field is as follows: Construct the following iterative equation: ΔP(k) I =P(k) I -P 实 (k) Among them, F(q) I+1 is the signal intensity of the qth valid voxel in the I+1th iteration, F(q) I is the signal intensity of the qth valid voxel in the Ith iteration, q = 1, 2, …, Q′, Q′ is the total number of valid voxels, α is the set relaxation factor, K′ is the total number of non-zero pixels contained in all two-dimensional projections, PSF is the mapping relationship matrix between each non-zero pixel and each valid voxel, ‖·‖ is the modulus operation, PSF(k,q) is the mapping value between the kth non-zero pixel and the qth valid voxel, ΔP(k) I is the simulated intensity P(k) obtained by the kth non-zero pixel in the Ith iteration I Compared with the actual strength P measured experimentally 实 (k) The error between Set the initial value F(q) for the signal intensity of each valid voxel 0 Then, the iteration equation is continuously iterated until the difference between the signal intensity of each effective voxel obtained by the iteration and the set convergence value is less than the set value, thereby completing the reconstruction of the light field distribution of the three-dimensional signal field.

2. A tomographic imaging method for overcoming internal light path obstruction as claimed in claim 1, characterized in that: The method for obtaining the spatial position coordinates of the obstruction in the discrete signal field to be measured is: If the design drawings of the three-dimensional signal field are unknown, multiple projections of the occluder in different directions are obtained by directly photographing the occluder, and then the occluder is reconstructed based on the projections to obtain its spatial position coordinates in the discrete signal field to be tested; if the design drawings of the three-dimensional signal field are known, the spatial position coordinates of the occluder in the discrete signal field to be tested are obtained by checking the design drawings.

3. A tomographic imaging method for overcoming internal light path obstruction as claimed in claim 1, characterized in that: After calibrating the spatial position and shooting direction of each image sensor, the calibrated image sensor is used to simultaneously obtain the two-dimensional projection of the three-dimensional signal field.

4. A tomographic imaging method for overcoming internal light path obstruction as claimed in any one of claims 1 to 3, characterized in that: The image sensor is a camera or a fiber bundle endoscope.

Images