A multi-cue fusion three-dimensional imaging method and system based on liquid lens array
By using time multiplexing technology with liquid lens arrays, multi-view, photometric stereo, and defocus depth images are acquired in time-division grouping, solving the trade-off between depth resolution and angular resolution in 3D imaging with solid microlens arrays, and achieving high-precision and fast 3D reconstruction results.
Patent Information
- Application Number
- CN202611629501.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing 3D imaging technologies based on solid microlens arrays suffer from trade-offs between depth resolution and angular resolution, with shallow depth of field leading to image blurring, high noise in depth maps, and blurred edges, making it difficult to achieve high-precision 3D reconstruction.
A liquid lens array is used for time multiplexing, and multi-view stereo, photometric stereo, and defocus depth images are acquired in time-division group. High-precision depth images are generated by multi-cue fusion. The electronically controlled zoom characteristics of the liquid lens array are used to achieve multi-view, multi-illumination, and multi-focal length image acquisition in a very small number of frames.
It achieves high-precision 3D reconstruction with sharp boundaries and rich high-frequency details, fast and mechanical-free operation, and is suitable for 3D imaging of complex targets. It is highly adaptable and applicable to fields such as industrial inspection and biomedical imaging.
Smart Images

Figure CN123118934A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of computational imaging and 3D vision technology, and in particular relates to a multi-cue fusion 3D imaging method and system based on a liquid lens array. Technical Background
[0002] Three-dimensional imaging has received continuous attention due to its wide range of applications in industrial precision inspection, micro-nano manufacturing, biomedical imaging, cultural heritage protection, and reverse engineering. Unlike long-range 3D imaging, which mainly focuses on the operating distance and detection sensitivity, the core technical indicators of short-range 3D imaging are concentrated on spatial resolution, depth accuracy, and measurement efficiency. Due to the extremely short imaging distance, the diffraction limit, depth of field limitation, and occlusion effect of the optical system are more prominent, and the complex reflection characteristics of the target surface (such as high reflectivity, translucency, and weak texture) also pose a serious challenge to the reconstruction quality.
[0003] A light field camera based on a solid-state microlens array is a typical solution for achieving single-frame 3D imaging. It provides a multi-view matching basis for multi-view stereo (MVS) by exposing multiple sub-lenses from different perspectives in a single exposure. Furthermore, its compact structure and lack of mechanical movement offer potential for miniaturization and high-speed measurement. However, this approach has significant drawbacks: the fixed focal length of the solid-state microlens array prevents true physical focal plane scanning, leading to a trade-off between depth-of-field resolution and angular resolution. Simultaneously, the small aperture and shallow depth of field inherent in the optical structure easily result in image blurring. In defocused conditions, the reliability of MVS matching decreases, depth map noise increases, edges become blurred, and even holes appear, making it difficult to provide reliable object boundaries. This limits the ability to achieve detailed and clearly defined 3D reconstructions. Summary of the Invention
[0004] To address the problems and needs existing in the background technology, this invention proposes a multi-cue fusion 3D imaging method and system based on a liquid lens array. This invention utilizes the time multiplexing of an actively electrically controlled liquid lens array to natively acquire images of three complementary cues within the same system. High-precision, sharp-boundary 3D depth recovery can be achieved with only a very small number of frames acquired, while avoiding the problems of large size, slow response, difficulty in integration, and real-time detection associated with mechanical scanning focusing or variable-angle devices. The method proposed in this invention is a multi-cue fusion 3D imaging method deeply coupled with the hardware characteristics of the liquid lens array.
[0005] The technical solution adopted in this invention is:
[0006] In a first aspect, the present invention proposes a multi-crucial fusion three-dimensional imaging method based on a liquid lens array, the method comprising the following steps:
[0007] S1: Using an imaging device containing a liquid lens array, multiple sets of images of the target under test are acquired in a time-division group. The multiple sets of images include a multi-view stereo (MVS) image group, a photometric stereo (PS) image group, and a depth-of-focus (DFD) image group. Each image in all image groups corresponds to the image of the target under test after passing through each liquid microlens, and each image is an RGB image.
[0008] S2: A first depth map is generated from the multi-view stereo image set, which represents the low-frequency shapes of the scene; a second depth map is generated from the photometric stereo image set, which contains high-frequency details; a boundary confidence map is generated from the defocus depth image set, which assigns higher weights to depth discontinuities than to flat areas; the multi-view stereo image set compensates for the lack of an absolute scale reference in the photometric stereo image set, while the photometric stereo image set compensates for the lack of detail in the multi-view stereo image set. The defocus depth image set has a strong response to depth discontinuity edges and is used to measure the weight of the fusion of the multi-view stereo image set and the photometric stereo image set.
[0009] S3: Generate the final depth map based on the first depth map, the second depth map, and the boundary confidence map.
[0010] Furthermore, the multi-view stereoscopic image group was obtained under uniform illumination and with all liquid microlenses in the liquid lens array having the same focal length.
[0011] Furthermore, the photometric stereo image group was obtained under non-uniform illumination and with all liquid microlenses in the liquid lens array having the same focal length, which is the same as the focal length of the liquid microlenses when acquiring the multi-view stereo image group.
[0012] Furthermore, the defocus depth image group was obtained under uniform illumination and with all liquid microlenses in the liquid lens array having different focal lengths.
[0013] Furthermore, the non-uniform lighting includes unidirectional lighting or bidirectional lighting.
[0014] Further, in step S2, generating a first depth map based on the multi-view stereo image group includes:
[0015] A multi-view stereo image group is processed using a multi-view stereo algorithm to obtain the first depth map.
[0016] Further, in step S2, generating a second depth map based on the photometric stereo image group includes:
[0017] A photometric stereo algorithm is used to process the photometric stereo image group to obtain a second depth map.
[0018] In step S2, generating a boundary confidence map based on the defocus depth image group includes:
[0019] The defocus response difference is obtained by calculating and normalizing the local defocus response differences between images at different defocus depths. The defocus response differences are expressed as Gaussian difference or Laplacian variance.
[0020] Further, S3 includes:
[0021] The low-frequency components of the first depth map and the high-frequency components of the second depth map are extracted. The extracted low-frequency components and high-frequency components are then superimposed to generate a composite depth map. The composite depth map is then subjected to edge-preserving filtering in conjunction with the boundary confidence map until optimization is complete, and the final depth map is obtained.
[0022] Furthermore, the superposition methods of the low-frequency components and high-frequency components include direct superposition and weighted superposition.
[0023] Furthermore, the edge-preserving filtering includes weighted least squares filtering, guided filtering, or bilateral filtering.
[0024] Secondly, this invention proposes a multi-crucial fusion three-dimensional imaging system based on a liquid lens array, the system comprising:
[0025] An imaging device containing a liquid lens array, used to acquire a set of images of the target under test;
[0026] A multi-directional illumination unit is used to illuminate the target being measured;
[0027] The control and computation processing unit is used to control the focal length of the liquid microlenses of the liquid lens array in the imaging device and the illumination mode of the illumination unit, as well as to generate the final depth map based on the acquired multi-view stereo image group, photometric stereo image group and defocus depth image group.
[0028] Furthermore, the liquid lens array includes multiple independently electrically controlled zoom liquid microlenses, positioned in front of the image sensor; the image sensor is used to receive the light beam passing through the liquid lens array and output an image.
[0029] Furthermore, the liquid lens array is an electrowetting lens array, with a number of microlenses not less than 3×3 and a spacing between sub-millimeter and millimeter levels.
[0030] Furthermore, the multi-directional lighting unit includes multiple independently controllable LED light sources arranged in a ring.
[0031] The beneficial effects of this invention include:
[0032] 1) High precision: This invention overcomes the shortcomings of a single method by using frequency domain division and fusion of three complementary clues: multi-view, multi-light, and multi-focus. The reconstructed depth map has the correct overall shape, rich high-frequency details, and sharp object boundaries.
[0033] 2) Fast and mechanically free: The electronically controlled zoom response of the liquid microlens reaches the millisecond level, and the illumination switching is synchronized with the image acquisition. The entire acquisition cycle can be controlled within tens of milliseconds. There is no macroscopic displacement mechanism, which is conducive to real-time detection and system miniaturization.
[0034] 3) High boundary fidelity: This invention specifically utilizes the high sensitivity of DFD to depth discontinuity to generate a boundary confidence map, which is used as a guide for edge preservation filtering, effectively solving the problems of object edge adhesion and contour blurring.
[0035] 4) High adaptability: The method of this invention has good reconstruction effect on samples with weak texture, high reflectivity or semi-transparentness, and can be widely used in circuit board solder joint detection, skin tissue imaging, microfluidic channel detection, industrial endoscopy and other fields. Attached Figure Description
[0036] Figure 1 This is a schematic diagram of the overall structure of the imaging system according to an embodiment of the present invention.
[0037] Figure 2 This is a schematic diagram of the structure of the target object being measured using an electrowetting liquid lens array according to an embodiment of the present invention.
[0038] Figure 3 This is a schematic diagram of patterned electrodes and coatings on the lower glass plate in an embodiment of the present invention; wherein, (a) is an electrode distribution diagram on the lower glass plate; and (b) is a schematic cross-sectional view of the coating on the lower glass plate.
[0039] Figure 4 This is an overall flowchart of the method of the present invention.
[0040] Figure 5 This is a schematic diagram of the multi-crucial original image group obtained by time-division group acquisition in an embodiment of the present invention; wherein, (a) is an MVS image group at different angles; (b) is a PS image group under different lighting conditions; and (c) is a DFD image group with different focal lengths.
[0041] Figure 6 This is a schematic diagram of the final generated 3D depth map of the die in an embodiment of the present invention.
[0042] In the diagram: 1. Target under test; 2. Multi-directional illumination unit; 3. Upper glass plate; 4. Conductive liquid; 5. Non-conductive liquid; 6. Lower glass plate; 61. Patterned electrode; 62. Copper wire; 63. Dielectric layer; 64. Hydrophobic thin film structure; 7. Electrowetting liquid lens array; 7-1. First liquid microlens; 7-2. Second liquid microlens; 7-3. Third liquid microlens; 7-4. Fourth liquid microlens; 7-5. Fifth liquid microlens; 7-6. Sixth liquid microlens; 7-7. Seventh liquid microlens; 7-8. Eighth liquid microlens; 7-9. Ninth liquid microlens; 8. Optical path mask; 9. Image sensor; 10. Imaging diffusion spot; 11. Control and computing processing unit. Detailed Implementation
[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. The described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.
[0044] The technical solution provided by the embodiments of the present invention will be described in detail below with reference to the accompanying drawings, taking the three-dimensional reconstruction process of a die as an example.
[0045] like Figure 1 As shown, the multi-crucial fusion three-dimensional imaging system based on a liquid lens array proposed in this invention includes:
[0046] An imaging device containing a liquid lens array, used to acquire a set of images of the target under test;
[0047] Multi-directional illumination unit 2 is used to illuminate the target being measured;
[0048] The control and calculation processing unit 11 is used to control the focal length of the liquid microlenses of the liquid lens array in the imaging device and the illumination mode of the illumination unit, as well as to generate the final depth map based on the acquired multi-view stereo image group, photometric stereo image group and defocus depth image group.
[0049] The imaging device of this embodiment includes an electrowetting liquid lens array 7 and an image sensor 9, with the image sensor 9 located behind the focal plane of the electrowetting liquid lens array 7. A multi-directional illumination unit 2 illuminates the target 1, and the light beam from the target 1 passes through the electrowetting liquid lens array 7 and then enters the image sensor. The image data acquired by the image sensor is transmitted to the control and computing processing unit 11. The target 1 is a die, located approximately 16 mm from the surface of the lens array.
[0050] In one feasible implementation, the electrowetting liquid lens array 7 includes an upper glass plate 3 and a lower glass plate 6 arranged in parallel and spaced intervals. The lower glass plate 6 is coated with a transparent patterned structure to achieve independent zooming of the liquid lens array. A conductive liquid 4 and a non-conductive liquid 5 are filled between the upper glass plate 3 and the lower glass plate 6. The conductive liquid 4 is disposed near the upper glass plate 3, and the non-conductive liquid 5 is disposed near the lower glass plate 6. A light path shield 8 is disposed on the side of the lower glass plate 6 away from the upper glass plate 3, and the light path shield 8 has several beam exits. The beam of the target object 1 passes through each liquid microlens and exits from the corresponding beam exit to form an imaging blur spot 10. Liquid microlenses with different focal lengths have their corresponding imaging blur spots 10.
[0051] The multi-directional lighting unit 2 is a ring-shaped LED array containing 8 LEDs. The 8 LEDs are evenly distributed on a hemisphere centered on the optical axis. Each LED can be switched on and off independently.
[0052] The electrowetting liquid lens array 7 employs a 3×3 liquid microlens structure, with a spacing of 2000 μm between adjacent microlenses. Each microlens can independently adjust its focal length within the range of 3.8-4.2 mm. For example... Figure 2 As shown, the nine liquid microlenses are respectively labeled as the first liquid microlens 7-1, the second liquid microlens 7-2, the third liquid microlens 7-3, the fourth liquid microlens 7-4, the fifth liquid microlens 7-5, the sixth liquid microlens 7-6, the seventh liquid microlens 7-7, the eighth liquid microlens 7-8, and the ninth liquid microlens 7-9.
[0053] In one feasible implementation, the image sensor 9 is a 10-megapixel CMOS sensor.
[0054] Figure 3 This is a schematic diagram of a liquid microlens structure capable of independent zoom. The coating structure on the lower glass plate 6 includes a patterned electrode 61. The patterned electrode 61 is connected to a copper wire 62, which is led out. A dielectric layer 63 is deposited above the lower glass plate 6 and the patterned electrode 61, and a hydrophilic film is coated on the surface of the dielectric layer 63. A patterned hydrophobic thin film structure 64, which is circular, is then deposited on the dielectric layer 63. The hydrophilic dielectric layer 63 and the patterned hydrophobic thin film structure 64 pin the hydrophobic non-conductive liquid 5 within the hydrophobic thin film structure 64, thereby forming an array of liquid microlenses. Optionally, the patterned electrode material is ITO, which is a transparent material. The dielectric layer 63 is made of SiO2. The hydrophobic thin film structure is made of Teflon.
[0055] like Figure 4 As shown, the multi-crucial fusion three-dimensional imaging method based on a liquid lens array proposed in this invention includes the following steps:
[0056] S1: Multiple sets of images of the target under test are acquired in a time-division and grouped manner using an imaging device containing a liquid lens array. These multiple sets of images include a multi-view stereo (MVS) image group, a photometric stereo (PS) image group, and a depth-of-focus (DFD) image group. Each virtual view image in each image group corresponds to the image of the target under test after passing through each liquid microlens. Each virtual view image is an RGB image, and the structure of a single image is as follows: Figure 5 As shown.
[0057] In one feasible implementation, the multi-view stereo image group is obtained by capturing images under uniform illumination and with all liquid microlenses in the liquid lens array having the same focal length. Specifically, all liquid microlenses are set to a focal length f0 = 4.0 mm, uniform illumination is turned on, and a light field original image is captured. After decoding the light field original image, pixels at the same position under each liquid microlens are extracted to generate 3×3 virtual view images, which are then used as the MVS image group.
[0058] In one feasible implementation, the photometric stereo image group is obtained under non-uniform illumination and with all liquid microlenses in the liquid lens array having the same focal length, which is the same as the focal length of the liquid microlenses when acquiring the multi-view stereo image group. Specifically, the focal length f0 is maintained, the illumination mode is changed to unidirectional illumination, and a photometric stereo image is obtained after capturing one frame and used as the PS image group.
[0059] In one feasible implementation, the defocus depth image group is acquired under uniform illumination and with all liquid microlenses in the liquid lens array having different focal lengths. Specifically, while maintaining uniform illumination, the focal lengths of the liquid microlenses surrounding the fifth liquid microlens 7-5 at the center of the electrowetting liquid lens array are rapidly adjusted to between 3.8mm and 4.2mm. After capturing one frame, an image in the defocus state is obtained, thus acquiring the DFD image group. The total acquisition time is approximately 50ms.
[0060] A schematic diagram of the multi-threaded original image obtained from the above steps is shown below. Figure 5 As shown, Figure 5 (a) is the MVS image group. Figure 5 (b) is a PS image group. Figure 5(c) represents the DFD image group. In the DFD image group, the focal lengths of the nine liquid microlenses are adjusted sequentially as follows: first liquid microlens 7-1, focal length 3.8 mm; second liquid microlens 7-2, focal length 3.9 mm; third liquid microlens 7-3, focal length 3.8 mm; fourth liquid microlens 7-4, focal length 4.1 mm; fifth liquid microlens 7-5, focal length 4.0 mm; sixth liquid microlens 7-6, focal length 3.9 mm; seventh liquid microlens 7-7, focal length 4.2 mm; eighth liquid microlens 7-8, focal length 4.1 mm; and ninth liquid microlens 7-6, focal length 4.2 mm. The adjustment of the liquid microlens focal lengths only requires that the depth of field covers the area of the object being measured.
[0061] S2: A first depth map is generated from the multi-view stereo image set, which represents the low-frequency shapes of the scene; a second depth map is generated from the photometric stereo image set, which contains high-frequency details; a boundary confidence map is generated from the defocus depth image set, which assigns higher weights to depth discontinuities than to flat areas; the multi-view stereo image set compensates for the lack of an absolute scale reference in the photometric stereo image set, while the photometric stereo image set compensates for the lack of detail in the multi-view stereo image set. The defocus depth image set has a strong response to depth discontinuity edges and is used to measure the weight of the fusion of the multi-view stereo image set and the photometric stereo image set.
[0062] In one feasible implementation, generating a first depth map based on a group of multi-view stereo images includes:
[0063] A multi-view stereo image set is processed using a multi-view stereo algorithm to obtain the first depth map. Specifically: for a 3×3 virtual view image, with the center view as a reference, a semi-global matching algorithm is used to calculate the disparity of each pixel and convert it into a depth map. This is recorded as the first depth map.
[0064] In one feasible implementation, generating a second depth map based on a set of photometric stereo images includes:
[0065] A photometric stereo image set is processed using a photometric stereo algorithm to obtain a second depth map. Specifically, the normal map is solved using the Lambert model and the known illumination direction, and then the relative depth is obtained by integration using the Frankot-Chellappa algorithm. This is recorded as the second depth map.
[0066] In one feasible implementation, generating a boundary confidence map based on the defocus depth image group includes:
[0067] The defocus response difference was calculated and normalized between images at different defocus depths. The defocus response difference was represented by either Gaussian difference or Laplacian variance. Specifically: for Figure 5 The defocus depth image group shown in (c) is divided into 9 groups of images with different focal lengths. For example, if the defocus response difference is expressed using Laplacian variance, then the Laplacian response is calculated within a 3×3 neighborhood of each pixel. The local variance is used as the defocus response of that pixel. Then, the difference in defocus response between images at different defocus depths was calculated. The formula is as follows:
[0068]
[0069] in, Indicates variance.
[0070] Finally, the differences in defocus response between images at different defocus depths were analyzed. Normalize to [0,1] to obtain the boundary confidence map. This can also be called a boundary probability heatmap. At points of abrupt depth changes in the defocus depth image, the boundary confidence map... Medium pixel The value at the location Approaching 1, in the flat areas of the defocus depth image, the boundary confidence map Medium pixel The value at the location It approaches 0. Therefore, The closer the value is to 1, the more likely that pixel is the boundary of the target being measured.
[0071] S3: Generate the final depth map based on the first depth map, the second depth map, and the boundary confidence map.
[0072] Specifically, it includes:
[0073] The low-frequency components of the first depth map and the high-frequency components of the second depth map are extracted. The extracted low-frequency components and high-frequency components are then directly superimposed to generate a composite depth map. The composite depth map is then subjected to edge-preserving filtering in conjunction with the boundary confidence map until optimization is complete, and the final depth map is obtained.
[0074] In one feasible implementation, a frequency threshold is set, with frequencies below the threshold designated as low frequencies and frequencies above the threshold designated as high frequencies. The first depth map can be low-pass filtered to obtain its low-frequency components; the second depth map can be filtered using a band-pass or high-pass filter to obtain its high-frequency components. For example, the cutoff frequency of the low-pass filter could be σ = 0.1 mm. -1 The passband of the bandpass filter is 0.1~2mm. -1 .
[0075] By combining the boundary confidence map, a weighted least squares (WLS) filter is used to perform edge-preserving filtering on the composite depth map. The optimization objective is to minimize the weighted sum of the data terms and the smoothing term, as shown in the following formula:
[0076] ,
[0077] Where p and q are pixel indices, and N(p) is the set of neighboring pixels of p. The depth used to constrain the output is close to the composite depth. Used to constrain the depth difference between adjacent pixels. This indicates a composite depth map generated by directly superimposing the extracted low-frequency and high-frequency components, with smoothing weights. From the boundary confidence map Modulation, specifically in the boundary confidence map Smoothing is suppressed in high-value areas and allowed in low-value areas. The final depth map is shown below. Figure 6 As shown.
[0078] Unlike traditional solid-state microlens arrays with fixed focal lengths or individual liquid microlenses that can only zoom as a whole, each microlens in a liquid lens array can have its focal length independently adjustable via voltage. The rapid zoom response characteristics and easily achievable independent focal length adjustment capability of electrowetting liquid microlenses effectively improve acquisition speed. The imaging device of this invention possesses array-level grouping rapid zoom capability, which is the hardware foundation for achieving rapid, mechanically motion-free, and high-fidelity 3D reconstruction. Therefore, this invention employs an independently electrically controllable zoomable electrowetting liquid lens array, enabling the simultaneous acquisition of multi-view, multi-illumination, and multi-focal-length images within a very small number of frames.
[0079] The scope of protection of this invention is not limited thereto. Any three-dimensional imaging scheme based on time multiplexing of active liquid lens arrays, which comprehensively utilizes the low-frequency information of MVS, the high-frequency information of PS, and the boundary guidance concept of DFD, is within the scope of this invention.
Claims
1. A multi-crucial fusion three-dimensional imaging method based on a liquid lens array, characterized in that, Includes the following steps: S1: Acquire multiple sets of images of the target under test using an imaging device containing a liquid lens array, wherein the multiple sets of images respectively include a multi-view stereo image set, a photometric stereo image set, and a defocus depth image set; S2: Generate a first depth map based on the multi-view stereo image group, generate a second depth map based on the photometric stereo image group, and generate a boundary confidence map based on the defocus depth image group; S3: Generate the final depth map based on the first depth map, the second depth map, and the boundary confidence map.
2. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 1, characterized in that, The multi-view stereoscopic image group was obtained under uniform illumination and with all liquid microlenses in the liquid lens array having the same focal length. The photometric stereo image group was obtained under non-uniform illumination and with all liquid microlenses in the liquid lens array having the same focal length, which is the same as the focal length of the liquid microlenses when acquiring the multi-view stereo image group. The defocus depth image group was obtained under uniform illumination and with all liquid microlenses in the liquid lens array having different focal lengths.
3. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 2, characterized in that, The non-uniform lighting includes unidirectional lighting or bidirectional lighting.
4. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 1, characterized in that, In step S2, generating a first depth map based on the multi-view stereo image group includes: A multi-view stereo image group is processed using a multi-view stereo algorithm to obtain the first depth map.
5. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 1, characterized in that, In step S2, generating a second depth map based on the photometric stereo image group includes: A photometric stereo algorithm is used to process the photometric stereo image group to obtain a second depth map.
6. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 1, characterized in that, S3 includes: The low-frequency components of the first depth map and the high-frequency components of the second depth map are extracted. The extracted low-frequency components and high-frequency components are then superimposed to generate a composite depth map. The composite depth map is then subjected to edge-preserving filtering in conjunction with the boundary confidence map until optimization is complete, and the final depth map is obtained.
7. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 6, characterized in that, The superposition methods of the low-frequency components and high-frequency components include direct superposition and weighted superposition.
8. The multi-crucial fusion three-dimensional imaging method based on a liquid lens array according to claim 6, characterized in that, The edge-preserving filtering includes weighted least squares filtering, guided filtering, or bilateral filtering.
9. A multi-crucial fusion three-dimensional imaging system based on a liquid lens array, characterized in that, include: An imaging device containing a liquid lens array, used to acquire a set of images of the target under test; A multi-directional illumination unit is used to illuminate the target being measured; The control and computation processing unit is used to control the focal length of the liquid microlenses of the liquid lens array in the imaging device and the illumination mode of the illumination unit, as well as to generate the final depth map based on the acquired multi-view stereo image group, photometric stereo image group and defocus depth image group.
10. A multi-crucial fusion three-dimensional imaging system based on a liquid lens array according to claim 9, characterized in that, The liquid lens array comprises multiple independently electrically controlled zoom liquid microlenses.
