Rapid three-dimensional shape measurement method for multi-scale frequency shift fringe projection

Abstract

The invention discloses a rapid three-dimensional shape measurement method for multi-scale frequency shift fringe projection. According to the method, on the basis of the method described in the basic patent CN118936361B, a frequency reuse stripe pattern is constructed through multi-scale region division, and images under different frequencies are obtained by adopting a three-step phase shift strategy. And constructing a complex signal for each sub-block image, performing frequency domain transformation, fusing main frequency amplitude images in two directions, extracting a frequency domain main peak index and an amplitude based on a position invariant theorem, realizing spatial positioning of a direct illumination main component, and further completing homonymy point matching and three-dimensional reconstruction between pixels of a camera and a projector. According to the method, only 15 patterns are needed in each direction, compared with 484 patterns in each direction in a basic scheme, the pattern compression rate reaches 96.9%, and the method is suitable for high-speed three-dimensional reconstruction tasks under complex illumination and dynamic scenes and has the advantages of being small in pattern number, high in response speed and high in positioning robustness.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional shape measurement technology, and in particular to a rapid three-dimensional shape measurement method using multi-scale frequency-shifting fringe projection suitable for complex lighting conditions (such as specular reflection, glossy surfaces, and translucent material scenarios). It can be widely applied in industrial scenarios such as precision manufacturing inspection, reverse engineering, and three-dimensional reconstruction of transparent / highly reflective objects. Background Technology

[0002] Fringe projection technology, due to its advantages of non-contact, high precision, and full-field measurement, has been widely used in precision manufacturing, reverse engineering, 3D reconstruction, and industrial inspection. Traditional fringe projection methods mainly include phase-shifting profilometry and Fourier profilometry. Their basic premise is that the projected pattern is modulated by the object's surface through a single diffuse reflection, and then received by a camera to form an image. However, in actual industrial scenarios, the objects being measured often have specular reflections, glossy surfaces, translucent materials, or multi-layered structures (such as metal parts, ceramic products, and glassware). These complex lighting conditions can cause multiple reflections, subsurface scattering, and ambient light interference, resulting in the image received by the camera being mixed with lighting information from multiple paths, thus severely interfering with phase extraction and 3D reconstruction. In this complex lighting propagation model, the fringe image received by the camera includes not only the one-time direct illumination component from the projector but also multi-path indirect illumination and background lighting. Spectral aliasing, modulation degradation, and phase distortion caused by complex reflections are among the core challenges limiting the widespread adoption of traditional fringe projection methods in industrial settings.

[0003] To eliminate the interference of strong mutual reflection on fringe projection 3D imaging, various improvement methods have been proposed. One approach uses high-frequency modulation patterns to suppress indirect illumination; however, on highly reflective surfaces, the fringe modulation level decreases significantly, easily leading to phase decoding failure. Other methods employ elastic modulation using high-frequency or low-frequency patterns to adapt to different lighting conditions, but these methods rely on a single light transmission model and struggle to handle complex light fields such as specular reflection or multiple scattering. Still other research introduces the single-pixel imaging principle to separate direct and indirect illumination, achieving high image quality; however, this typically requires projecting tens of thousands of patterns, resulting in high data acquisition and processing costs, making it difficult to meet the requirements of real-time or high-speed measurements.

[0004] To further improve adaptability to complex lighting conditions, some researchers have combined frequency analysis with environmental matting techniques to propose methods for 3D reconstruction of transparent or glossy objects. These methods extract direct illumination components through static background modeling, but the spectral response becomes blurred in areas with high curvature or bright spots, easily leading to 3D reconstruction failure. Building on this, the Baseline Frequency Shift Method (BFSM) was proposed. This method projects multiple sets of sinusoidal fringe patterns at different frequencies, frequency-encodes the projector rows and columns, and uses Discrete Fourier Transform to extract the main peak positions of each frequency response, establishing a mapping relationship between camera pixels and projection coordinates. While BFSM possesses some anti-interference capabilities, it relies on projecting a large number of structured light patterns, resulting in low system efficiency. Furthermore, the main frequency extraction process depends on the precise location of local peaks in the spectrum, making it susceptible to background light disturbances and spectral noise, posing a risk of misselecting the main peak and causing reconstruction errors.

[0005] To address the aforementioned issues, patent CN118936361B proposes a method for measuring the 3D shape of frequency-shifted fringes under complex lighting conditions. This method employs multi-frequency patterns and a four-step phase-shift strategy, combined with epipolar constraints for dominant frequency extraction. However, this approach still suffers from the following problems: ① The fixed-frequency encoding method results in low pattern utilization, requiring the projection of a large number of patterns (up to 484 patterns per direction), limiting the system's frame rate and measurement efficiency; ② Epipolar constraints rely on spectral continuity, making them prone to mis-extraction in high-curvature or blurred-edge regions; ③ The four-step phase-shift method leads to a large data volume and long acquisition and processing time, making it unsuitable for dynamic scenes. Therefore, an improved frequency domain measurement method with fewer patterns, more robust dominant frequency positioning, and a more robust decoding path is urgently needed to maintain anti-interference capabilities while adapting to the frame rate and efficiency requirements of high-speed 3D reconstruction scenarios. The technical solution of this invention to solve the above problems is: a rapid 3D shape measurement method based on multi-scale frequency-shifted fringe projection, comprising the following steps:

[0006] Step 1: The key feature is the adoption of a multi-scale region partitioning strategy, dividing the image into s×s sub-blocks. Each sub-block embeds a three-step phase-shift sinusoidal fringe pattern with frequencies of 8, 12, 16, 20, and 24 through pattern scaling and replication, constructing a spatial frequency reuse structure. This design differs from traditional schemes where each frequency group requires independent projection, achieving a multi-frequency nested structure within the same pattern, improving frequency coding density, and significantly reducing the number of required patterns.

[0007] Step 2: Reconstruct the complex signal using the three-step phase-shifting method based on the preset multi-scale image blocks, and obtain the local frequency response coefficients using Fourier transform.

[0008] Step 3: The key feature is that frequency response maps are constructed in both the horizontal and vertical directions, and amplitude-weighted fusion is performed to generate a two-dimensional master response map (T_block). Based on the characteristic that the spatial position of the master frequency response remains unchanged in each image block, a unified frequency domain template is used to perform fast master peak localization on T_block, extracting the master frequency index and response amplitude, thereby achieving robust identification and projection spatial localization of the direct illumination master component.

[0009] Step 4: Obtain corresponding points between projector and camera pixels by directly illuminating the location, and generate a depth map based on triangulation.

[0010] The above-mentioned fast three-dimensional shape measurement method based on multi-scale frequency-shifting fringe projection, specifically step one, is as follows:

[0011] 1-1) Unlike traditional unified frequency coding schemes, this invention is based on a multi-scale image block partitioning strategy. Multiple local regions are defined within the entire projected image, and each region embeds a three-step phase-shifted sinusoidal fringe pattern using a frequency reuse nesting method. The selected frequency components include 8, 12, 16, 20, and 24, which are used to construct horizontal and vertical bidirectional frequency coding structures, respectively. This allows different image blocks to carry independent frequency information in the spatial domain, achieving high-density frequency coding and pattern number compression.

[0012] 1-2) Based on the selected frequency and phase shift step, a stripe pattern matching the sub-region frequency encoding is automatically generated. Each pattern adopts a three-step phase shift strategy, constructs horizontal and vertical sinusoidal stripe patterns through a frequency-region mapping table, and achieves multiplexing of multiple frequencies within the region through pattern scaling and spatial distribution, thereby improving frequency utilization and pattern expression efficiency.

[0013] 1-3) By applying equally spaced phase shifts to the stripe pattern, such as A three-step phase-shift sequence is formed. This process is completed by the control software dynamically indexing the pattern sequence, without the need for physical displacement or background image superposition, and can be directly used for subsequent complex signal construction.

[0014] 1-4) The computer-generated horizontal and vertical stripe images are burned into the projector. By precisely controlling the projector, the stripe images are made to evenly cover the surface of the object, and the horizontal and vertical stripe patterns are captured using an area scan camera.

[0015] The above-mentioned fast three-dimensional shape measurement method based on multi-scale frequency-shifting fringe projection, specifically step two, is as follows:

[0016] 2-1) Constructing complex signals based on three-step phase-shifted image sequences, specifically using the following formula:

[0017]

[0018] This expression constructs a complex signal by the difference between phase diagrams, effectively suppressing the influence of DC terms and background light, and is suitable for extracting the main peak of the spectrum in frequency reuse structures. To improve the stability of frequency response extraction, the above complex signal construction method C(x,y) is used to concentrate the spectral energy at the specified main frequency position.

[0019] 2-2) Perform Discrete Fourier Transform (DFT) on the complex signal to obtain the frequency domain signal corresponding to each image block.

[0020] 2-3) In the frequency domain, only the response components corresponding to the projected frequency in the positive spectrum are retained, while the DC component and the negative spectrum term are removed, thereby effectively suppressing background illumination interference.

[0021] 2-4) Based on the spectral analysis of the direct and indirect illumination components, accurately distinguish the main light sources and reflected light affecting imaging. Detailed analysis of each peak in the frequency domain signal effectively suppresses interference from background illumination, ensuring the accuracy of the spectral analysis.

[0022] The above-mentioned rapid three-dimensional shape measurement method based on multi-scale frequency-shifting fringe projection, specifically step three is as follows:

[0023] 3-1) For the acquired multi-frequency three-step phase-shifted images, complex signals are constructed according to the preset image block division, and a Discrete Fourier Transform (DFT) is performed on each image block to calculate the local frequency response maps in the horizontal and vertical directions respectively. Then, the amplitudes of the corresponding frequency response maps in the two directions are summed and fused to generate a two-dimensional master response map.

[0024] 3-2) Search for the maximum frequency response point in each image block of the two-dimensional master response map, and extract the position and amplitude of the master frequency peak. The frequency point corresponding to the master response peak represents the frequency domain position of the principal component of direct illumination received by the current camera pixel on the projector image plane.

[0025] 3-3) Based on the consistency of the structural replication of the multi-frequency stripe pattern in each image block, its main frequency response position in the frequency domain remains relatively stable. A unified frequency domain template can be used to quickly locate the main response peak, thereby improving the stability and computational efficiency of the main frequency recognition.

[0026] 3-4) Based on the extracted main response frequency position and its amplitude, a pixel-level mask is constructed to remove image blocks with the main peak amplitude below the set threshold, effectively suppressing background interference and indirect illumination, and improving the frequency domain response stability and subsequent three-dimensional reconstruction accuracy.

[0027] The above-mentioned rapid three-dimensional shape measurement method based on multi-scale frequency-shifting fringe projection, specifically step four, is as follows:

[0028] 4-1) The system creates the corresponding projection matrix by loading the calibrated camera and projector parameters.

[0029] 4-2) Based on the coordinates of the corresponding points and the known geometric relationship between the camera and the projector, calculate the three-dimensional coordinates and convert the calculated depth value Z into a depth map.

[0030] 4-3) Extract 3D coordinates from the depth map to generate point cloud data.

[0031] To further clarify the technical differences between this invention and the basic patent CN118936361B, and the basis for not falling within its scope, please refer to Table 1 below. Table 1 lists the differences between the basic patent and this invention in key technical dimensions such as stripe structure design and frequency processing strategy, and serves as the basis for supporting the argument that this invention does not fall within the scope of protection of the basic patent claims.

[0032] The beneficial effects of this invention are as follows:

[0033] 1. This invention, through pattern scaling and replication and multi-frequency nesting design, embeds multiple frequency components in each image sub-block, constructing a spatial frequency reuse structure within the region. This method significantly improves frequency coding density and avoids the redundant design of traditional methods where "each group of frequencies needs to be encoded separately." Figure 4 As shown, this invention only requires projecting 15 patterns in each direction to complete the full frequency encoding, compared to the basic patent which requires projecting 484 patterns in each direction. The pattern number compression ratio is as high as 96.9%. Under the same hardware conditions, the processing time for each complete 3D reconstructed image of this invention is no more than 30 seconds, while the processing time for the basic patent solution under the same conditions is more than 300 seconds. The overall processing efficiency is improved by about 90%, and the image acquisition and decoding cycle of the system is significantly shortened, making it more suitable for high-speed 3D measurement needs in dynamic scenes.

[0034] 2. This invention directly extracts the dominant frequency response position in the frequency domain, replacing the phase unfolding step in the traditional phase-shifting method, thus avoiding the problem of phase distortion accumulation under complex lighting conditions such as strong background light and specular reflection. For example... Figure 6 As shown, the main peak of the frequency response exhibits a stable and concentrated distribution in each image sub-block, supporting rapid matching and localization using a unified frequency domain template. This scheme requires no background modeling or epipolar constraints and still maintains good 3D reconstruction accuracy and robustness in highly interference scenes such as glossy surfaces and occluded areas. Attached Figure Description

[0035] Figure 1 This is a schematic diagram of the main optical structure used in this invention.

[0036] Figure 2 This is a flowchart of the detection process of the present invention.

[0037] Figure 3 This is a schematic diagram illustrating the principle of multi-frequency Fourier stripe pattern generation in this invention.

[0038] Figure 4 The frequency reuse fringe pattern (s=6,f=[8-24]) is constructed based on multi-scale partitioning in this invention.

[0039] Figure 5 The images are obtained using an area scan camera in this invention.

[0040] Figure 6 This is a diagram showing the amplitude distribution of the main frequency response in this invention.

[0041] Figure 7 This is a three-dimensional reconstruction image of the object being tested according to the present invention. Detailed Implementation

[0042] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0043] 1. For example Figure 1 As shown, the projector is used to load and project multi-frequency sinusoidal fringe patterns. Its resolution and frame rate determine the spatial accuracy and temporal response capability of the measurement system. To meet the needs of high-speed measurement scenarios, a projection device with a high frame rate and matching resolution should be selected. The working distance should be determined according to the actual size of the object being measured, and it is recommended to keep it between 0.5m and 1.0m.

[0044] 2. Area scan camera: To ensure the acquisition of complete fringe patterns and obtain high-quality frequency response, this invention selects an area scan camera with a resolution of 2448×2048, which matches the resolution of the projector and meets the requirements of high-speed measurement.

[0045] 3. To reduce image distortion caused by the undulation of objects, this invention uses a telecentric lens instead of a conventional lens, effectively suppressing the problem of near objects appearing larger than distant objects caused by changes in the imaging angle, and improving structural consistency and 3D reconstruction accuracy.

[0046] like Figure 2 As shown, a fast three-dimensional shape measurement method using multi-scale frequency-shifting fringe projection includes the following steps:

[0047] Step 1: Based on the multi-scale region division strategy, determine the frequency range and generate and project three-step phase-shift fringe patterns at different frequencies.

[0048] The specific process of step one is as follows:

[0049] 1-1) As Figure 3As shown, multiple coding frequencies (e.g., 8, 12, 16, 20, 24) are selected. The selected frequency components are determined based on the following principles: ① adjacent frequency intervals ≥ 4 to avoid spectral aliasing; ② frequency range covering 8-24 to suit the local spectral resolution requirements of sub-block size (s×s); ③ all are positive integer frequencies to facilitate peak localization after Fourier transform. A three-step phase-shifting strategy is combined to construct a multi-frequency multiplexed Fourier fringe pattern. The fringe pattern is scaled and copied to multiple sub-regions in the image at a fixed ratio, thereby enhancing local spectral resolution and improving the accuracy of main frequency peak extraction.

[0050] 1-2) such as Figure 4 As shown, a three-step phase-shift strategy is used to encode each group of frequencies, processing a total of 5 × 3 = 15 images. The above pattern is based on... Figure 3 The multi-scale partitioning principle illustrated in the experiment sets the number of sub-regions to s=6. This setting of s=6 is based on a balance between projector resolution and local spectral analysis accuracy: if s is too small, the sub-blocks are too large, resulting in blurred local frequency responses; if s is too large, the sub-blocks are too small, leading to dispersed spectral energy. Experiments have verified that s=6 can balance accuracy and efficiency. Specifically, the stripe pattern is scaled and replicated in a 6×6 sub-region structure within the image, and the resulting horizontal and vertical stripe images of different frequencies are then burned into the projector.

[0051] 1-3) such as Figure 5 As shown, the original image is acquired by an area scan camera. The use of a telecentric lens can overcome the phenomenon of near objects appearing larger than distant objects due to height differences, thereby reducing imaging distortion.

[0052] Step 2: Reconstruct the complex signal using the three-step phase-shifting method based on the preset multi-scale image blocks, and obtain the local frequency response coefficients using Fourier transform.

[0053] Step two involves the following steps:

[0054] 2-1) For each acquired stripe image, it is divided into multiple image blocks for local processing. The stripe pattern within each image block can be regarded as a real signal, and the real signal spectrum response obtained from the stripe pattern is obtained.

[0055]

[0056] Among them, C DC For the DC component at position c p The spectral response generated by =0, for real signals will It will produce conjugate spectral responses at two locations, with response intensities of The negative spectrum portion is redundant.

[0057] 2-2) To remove the DC component and negative spectral response, a three-step phase-shifting process is performed on the fringe pattern projected onto the target object. Specifically, three fringe patterns with different phases are collected by changing the phase (0°, 120°, 240°) of each image. Let... Substitute into the following formula. Let the three images be I0, I1, and I2, respectively.

[0058]

[0059] 2-3) Calculate the difference: Use these three images to calculate two key difference values:

[0060] This step calculates the difference of the cosine terms of the phase-shifted image, providing the real part of the complex signal.

[0061] This step calculates the difference of the sinusoidal terms of the phase-shifted image, providing the imaginary part of the complex signal.

[0062] 2-4) The real part D obtained through the above calculation C and the imaginary part D S Complex signals can be synthesized:

[0063] D = D C +jD S

[0064] The obtained complex signal D is subjected to Fourier transform in each image block to obtain the signal's spectral response C. In frequency domain processing, the system retains only the positive frequency components, filters out the DC component and the mirror negative frequency response, effectively suppresses unstructured light interference, improves the clarity of the main frequency energy peak, and provides robust spectral support for subsequent main frequency point extraction.

[0065] 2-5) Perform amplitude analysis on the spectral response C within each image block, extracting the location and intensity of the dominant frequency response with the largest amplitude in both the horizontal and vertical directions to obtain the dominant frequency response distribution of the entire image under multi-frequency interference. For example... Figure 6 As shown, the dominant frequency response curve extracted from multi-scale image patches in this invention exhibits multiple clearly distinguishable spectral peaks. Combined with a unified frequency domain template, it can quickly and stably locate the dominant frequency point. Compared to the strong dependence of basic patented methods on spectral epipolar lines, this method is more robust to calibration errors, especially exhibiting stronger dominant frequency recognition capabilities in blurred edge regions.

[0066] Step 3: Merge the frequency response amplitude maps from both directions, and based on the position invariance theorem, search for the point with the maximum spectral energy to extract the dominant frequency index and response amplitude, thereby achieving the localization of the direct illumination position.

[0067] Step three involves the following steps:

[0068] 3-1) For the acquired multi-frequency three-step phase-shifted images, complex signals are constructed according to the preset image block division, and a Discrete Fourier Transform (DFT) is performed within each image block to calculate its local frequency response map in the horizontal and vertical directions. Subsequently, the amplitudes of the frequency response maps in the two directions are weighted or equally weighted and fused to construct a two-dimensional principal frequency response map of the corresponding region, which is used to enhance the significance of the principal frequency components.

[0069] 3-2) Search for the point with the maximum spectral energy in each image block of the two-dimensional master response map, and extract the position index and amplitude of the master frequency peak in that block. The master response peak represents the frequency domain coordinates of the direct illumination master frequency component received by that image block on the projected image plane.

[0070] 3-3) Based on the position invariance theorem, since the structural replication of the multi-frequency stripe pattern in each image block is consistent, the relative position of the main frequency in the frequency domain remains unchanged. Thus, the main response peak can be quickly located in a unified frequency domain template, improving the stability and computational efficiency of main frequency recognition.

[0071] 3-4) A pixel-level mask is constructed based on the extracted principal response frequency position and its amplitude to filter regions with reliable frequency domain response. Image blocks with principal frequency amplitudes below a set threshold are removed, effectively suppressing the impact of background interference, indirect illumination, and edge spectral aliasing on reconstruction accuracy. Finally, within the effective mask range, the normalized principal frequency index position is mapped to the projector coordinate system to obtain a disparity map, providing input for 3D point cloud computing and spatial geometric reconstruction.

[0072] Step 4: Obtain corresponding points between projector and camera pixels by directly illuminating the location, and generate a depth map based on triangulation.

[0073] Step four involves the following steps:

[0074] 4-1) Load the configuration data of the calibrated camera and projector to construct a matrix for image projection.

[0075] 4-2) such as Figure 7 As shown, the system determines the coordinates of a unique corresponding point in the camera pixel and projector coordinate system based on the dominant frequency index position. This corresponding point is uniquely determined by the position corresponding to the energy peak in the dominant frequency response diagram, eliminating the need for the multi-step calculation process based on phase unrolling and anti-wrapping paths in traditional methods. By directly establishing the spatial pixel mapping relationship between the projector and camera in the frequency domain, the system significantly reduces error propagation caused by phase jumps and mismatches, thereby improving the stability and accuracy of 3D reconstruction.

[0076] 4-3) Save the processed point cloud in a standard format for further analysis or visualization to complete high-precision 3D measurement.

[0077] Table 1 Comparison of Improvements and Basic Patents

[0078]

[0079] As shown in Table 1, compared with the fixed frequency, four-step phase shift, and epipolar constraint triangulation measurement process disclosed in the basic patent CN118936361B, the present invention has significant structural differences and performance improvements in terms of pattern quantity, acquisition efficiency, and spectrum positioning method by constructing multi-frequency nested patterns, three-step phase shift complex signal reconstruction and main response map peak extraction method, and does not fall within the protection scope of the basic patent.

Claims

1. A rapid three-dimensional shape measurement method based on multi-scale frequency-shifting fringe projection (referencing basic patent CN118936361B), characterized in that, Includes the following steps: Step 1: Based on the multi-scale region segmentation strategy, embed multiple frequency (three-step phase-shift coding) stripe patterns in each image sub-block. The patterns are scaled and copied to construct a frequency reuse structure. Step 2: Perform three-step phase-shift decoding on the acquired image to construct a complex signal and obtain the frequency response amplitude distribution of each sub-block; Step 3: Fuse the amplitude maps of the main frequency response in the horizontal and vertical directions, extract the main peak index and amplitude based on the spatial position invariance of the main frequency, and realize the localization of the main component of direct illumination; Step 4: Obtain corresponding points between projector and camera pixels by directly illuminating the location, and generate a depth map based on triangulation.

2. The rapid three-dimensional shape measurement method based on multi-scale frequency-shifting fringe projection according to claim 1, characterized in that, The specific process of step one includes: 1-1) The stripe pattern adopts a multi-scale region division method, embedding stripe patterns with frequencies of 8, 12, 16, 20 and 24 in each image sub-block, and constructing a frequency reuse structure by combining a three-step phase shift and pattern scaling and copying strategy. 1-2) Generate a periodic stripe pattern using a computer. Calculate the intensity variation of each stripe in different phases according to the following formula. This process includes calculating the stripes at φ = 0, The intensities of the three phases are used to generate corresponding stripe images. 1-3) Burn the pre-generated horizontal and vertical stripe images into the projector and project them so that the images completely cover the object to be tested. 1-4) Use a high-resolution area array camera with a telecentric lens to acquire images to reduce distortion caused by differences in imaging distance.

3. The rapid three-dimensional shape measurement method using multi-scale frequency-shifting fringe projection according to claim 2, characterized in that, The specific process of step two includes: 2-1) The acquired fringe image is divided into regions according to a preset multi-scale image block division method. Within each image block, a three-step phase shift operation is performed. Based on the intensity differences of the fringe pattern at different phases, the difference between the real and imaginary parts is calculated using phase difference to construct a complex signal, where I0, I1, and I2 represent the values ​​at phase shifts of 0, 1, 2, and 3 respectively. and The intensity values ​​of the striped pattern collected under the given conditions. This step calculates the difference of the cosine terms of the phase-shifted image, providing the real part of the complex signal. This step calculates the difference of the sinusoidal terms of the phase-shifted image, providing the imaginary part of the complex signal. Complex signal: D = D C +jD S 2-2) Perform a Discrete Fourier Transform (DFT) on the complex signal obtained above to obtain the frequency domain signal. Only retain the positive spectral component corresponding to the target frequency for the calculation of the main peak response, ignoring low-frequency DC and high-frequency noise. The response intensity at each frequency can be extracted through frequency domain analysis, and subsequent multi-frequency fusion and main peak localization operations can be performed to improve the reconstruction robustness and accuracy under complex lighting conditions.

4. The rapid three-dimensional shape measurement method using multi-scale frequency-shifting fringe projection according to claim 3, characterized in that, The specific process of step three is as follows: 3-1) For the acquired multi-frequency three-step phase-shift images, Fourier transform analysis is performed according to the preset image block division. Local frequency response maps in the horizontal and vertical directions are constructed in each image block, and the corresponding block regions are summed and fused to form a two-dimensional master response map. 3-2) Search for the maximum frequency response point in each image block of the two-dimensional master response map, and extract the peak position and amplitude of the master frequency. The peak point corresponds to the spatial position of the principal component of direct illumination received by the camera pixel on the projector image plane. 3-3) Based on the consistency of the structural replication of the multi-frequency stripe pattern in each image block, the relative position of the main frequency in the frequency domain remains unchanged, so that the main response peak can be quickly located in a unified frequency domain template, thereby improving the stability and computational efficiency of the main frequency recognition.

5. The rapid three-dimensional shape measurement method using multi-scale frequency shift fringe projection according to claim 4, characterized in that, The specific process of step four is as follows: Based on the projector image coordinates corresponding to the main peak in the frequency domain, and combined with the joint calibration parameters of the camera and the projector, the corresponding point relationship between the pixels at the imaging end and the projection end is determined; a triangulation model is used to calculate the three-dimensional coordinates of each pair of pixels, constructing a depth map and a three-dimensional point cloud, thereby achieving high-precision three-dimensional reconstruction under complex lighting conditions.

Images