Method for binary fringe projection profilometry based on improved error diffusion

Abstract

The present application relates to the technical field of optical three-dimensional profile measurement, and particularly relates to a binary fringe projection profilometry method based on improved error diffusion, which comprises the following steps: firstly, generating a plurality of ideal sinusoidal fringe images; using an improved error diffusion algorithm to binarize the ideal sinusoidal fringe images into 1-bit fringe patterns, wherein the algorithm improves the sinusoidal property of the patterns by using Sierra-Lite weights, introducing random perturbation and snake scanning sequence; then projecting the binary fringe patterns to the surface of a measured object under the focusing state of a projector, and collecting deformed fringe images; finally, extracting phase based on a multi-step phase shifting method and completing three-dimensional reconstruction. The present application breaks through the technical limitation of traditional binary fringe projection which relies on defocusing operation, and can realize high-precision and high-speed three-dimensional measurement under the condition of complete focusing. The present application has the advantages of strong robustness, simple system construction, good hardware compatibility and the like, and is suitable for various application scenarios such as industrial detection and dynamic profile analysis.

Description

Technical Field

[0001] This invention relates to the field of optical three-dimensional topography measurement technology, and in particular to a binary fringe projection profilometry method based on improved error diffusion. Background Technology

[0002] Binary Fringe Projection Profilometry (BFPP) is an important research direction in optical measurement, playing a crucial role in modern manufacturing and measurement technology. Unlike traditional fringe projection techniques that rely on ideal 8-bit sinusoidal fringes, BFPP uses a 1-bit quasi-sinusoidal fringe pattern, avoiding gamma calibration in commercial projectors and supporting high-speed projection exceeding 4kHz. Existing BFPP methods (also known as binary defocusing methods) are mainly based on one-dimensional fringe patterns generated by square wave fringes combined with various pulse width modulation strategies, and two-dimensional fringe patterns generated using traditional halftone algorithms, such as the Bayer dithering algorithm and the Floyd-Steinberg error diffusion algorithm (FSED). All of these methods require defocusing the projection optical system to blur the projected binary fringes during imaging, thereby suppressing higher-order harmonic components and improving its approximation to the ideal sinusoidal fringes.

[0003] However, projection defocusing has significant limitations in practical applications. First, the degree of defocus is highly dependent on hardware configuration, and the adjustment process requires manual intervention, making it difficult to accurately reproduce. There is a lack of unified defocus quantification standards across different projection devices, so "weak," "medium," and "strong" defocus are merely subjective. Second, structured light measurement systems generally rely on clearly focused projection fringes to ensure calibration accuracy; defocusing disrupts this process and reduces the efficiency of the measurement system. Third, excessively high defocus levels significantly weaken fringe contrast, leading to phase extraction failure. Fourth, defocused fringes reduce the effective depth of field for measurement.

[0004] Generally, one-dimensional patterns are more sensitive to defocusing than two-dimensional patterns because two-dimensional patterns generated by halftone methods are inherently closer to sinusoidal waveforms, thus having less dependence on defocusing. Zhou et al. demonstrated that under approximate focusing conditions, patterns generated using FSED with dynamic thresholding, as well as patterns generated by combining FSED with spatiotemporal coding, can achieve accurate 3D reconstruction. This prompts us to explore a possibility: eliminating the step of manually defocusing the projector in all previous BFPP methods, and instead projecting a 1-bit pattern generated by a high-quality two-dimensional halftone method to better approximate a sinusoidal wave under focused projection, thereby breaking the traditional binary defocusing paradigm and solving the technical bottleneck caused by manual defocusing. Therefore, it is necessary to propose a binary fringe projection method that does not require defocusing, has high fringe quality, and excellent measurement accuracy. Summary of the Invention

[0005] The purpose of this invention is to provide a binary fringe projection profilometry method based on improved error diffusion, so as to solve the problem of dependence on projection defocus in traditional methods and realize high-speed and high-precision three-dimensional measurement in the focused state.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: This invention provides a binary fringe projection profilometry method based on improved error diffusion, comprising the following steps: S1. Obtain multiple ideal stripe images with sinusoidal grayscale distribution, and generate corresponding 1-bit binary stripe patterns based on the improved error diffusion algorithm; The improved error diffusion algorithm includes using the Sierra-Lite error diffusion weight matrix for error allocation; introducing random perturbations during weight allocation; and traversing image pixels in a serpentine scanning order. S2. With the projector focused, the binary stripe pattern is sequentially projected onto the surface of the object being measured; and the deformed stripe image sequence modulated by the surface of the object being measured is acquired. S3. Extract phase information based on the deformed stripe image sequence and perform three-dimensional morphology reconstruction.

[0007] Furthermore, the random disturbances are grouped according to their weights, and the amplitude of the random disturbance in each group does not exceed 50% of the smallest weight in that group.

[0008] Furthermore, the serpentine scanning sequence is to scan even-numbered rows of the image from left to right and odd-numbered rows from right to left.

[0009] Furthermore, the focused state refers to the projector's optical system being in the state of optimal imaging clarity, without the need for additional defocusing operations.

[0010] Further, phase information is extracted based on the deformed stripe image sequence, and three-dimensional topography reconstruction is performed, including: The wrapping phase is calculated from the deformed stripe image sequence based on the multi-step phase shift method; The absolute phase distribution is obtained by performing phase unrolling on the wrapped phase. Based on the system calibration parameters, the absolute phase is mapped to three-dimensional spatial coordinates.

[0011] Furthermore, the multi-step phase-shifting method has at least 3 phase-shifting steps.

[0012] Furthermore, the projection frame rate of the binary stripe pattern is not less than 1000Hz, and it has a quasi-sinusoidal light intensity distribution under focused projection.

[0013] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention breaks through the traditional binary fringe projection method, which relies on projector defocusing. Existing methods generally require manual defocusing to suppress higher harmonics of binary fringes in order to approximate ideal sinusoidal fringes. However, the degree of defocusing is difficult to control and reproduce precisely, and it introduces additional operational complexity and uncertainty. This invention performs measurements with the projector fully focused and generates binary fringes with high sinusoidality through the proposed improved error diffusion algorithm. It achieves high-quality fringe projection contouring without any defocusing adjustment, eliminating problems such as calibration interference, decreased fringe contrast, decreased effective measurement depth of field, and poor consistency between devices caused by defocusing.

[0014] (2) Experimental data show that, under focused conditions, the measurement accuracy of this invention is optimal, with the root mean square error of phase measurement as low as 0.0012 rad and the root mean square error of height reconstruction as low as 0.0884 mm. Its accuracy is significantly better than the benchmark method using the traditional Floyd-Steinberg error diffusion algorithm. Furthermore, this invention exhibits strong robustness to defocus conditions. Under weak, medium, and strong defocus conditions, the mean and standard deviation of its phase and height measurement errors are significantly lower than those of the benchmark method. This ensures that even if the system experiences slight defocusing due to vibration, thermal drift, or calibration deviation in practical applications, it can still maintain stable and reliable measurement results, significantly improving the system's practicality and environmental adaptability.

[0015] (3) Since the generated and projected pattern is a 1-bit binary image, its data transmission and processing load is extremely small, enabling the digital light processing projector to project at an extremely high refresh rate of over 4 kHz, which is far higher than the projection frame rate of traditional 8-bit grayscale sinusoidal stripes. Combined with synchronous high frame rate camera acquisition, high-speed 3D reconstruction can be easily achieved, providing a powerful tool for efficient online detection in production lines and morphological measurement of moving objects.

[0016] (4) No need for tedious and subjective defocus adjustment steps, making the system calibration process more direct and shortening the system deployment time. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the method flow of the present invention.

[0018] Figure 2 This study presents the 3D reconstruction results obtained using different binary fringes and different defocus states under the conditions of a projection fringe period of 10 pixels and 4-step phase shift, as well as the height error analysis of each group of reconstruction results in column 3800.

[0019] Figure 3 This shows the contour comparison of column 3800 in each group of experiments. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] like Figure 1 As shown, this embodiment discloses a binary fringe projection profilometry method based on improved error diffusion, which includes the following steps: Step 1: Obtain multiple ideal stripe images with sinusoidal grayscale distribution, and generate corresponding 1-bit binary stripe patterns based on the improved error diffusion algorithm; In this embodiment, multiple 8-bit grayscale sinusoidal stripe images are generated based on the phase shift step number N and the stripe width; then, a corresponding 1-bit binary stripe pattern is generated based on the improved error diffusion algorithm.

[0022] This embodiment directly improves the error diffusion algorithm through the following steps to enhance the quality of halftone stripe patterns.

[0023] First, the weights of the traditional Floyd-Steinberg error spread filter are... Replace with Sierra-Lite weights .

[0024] Secondly, drawing inspiration from Ulichney's method, random weights are introduced into the error diffusion filter. This strategy has been proven effective in improving halftone pattern quality in both the spatial and frequency domains. Specifically, two sets of larger weights are paired, and smaller weights are assigned to another set. A random variable is introduced into each set of weights, with an amplitude half that of the smaller weight in that set. The final error diffusion filter weights are then determined. ,in , Representing an interval A continuous and uniform distribution within.

[0025] Finally, the traditional row-by-row left-to-right binary pixel generation order is replaced with a serpentine order to improve the radial symmetry of the output halftone pattern and reduce directional artifacts. For each pixel, its normalized gray value is added to the accumulated error from the neighborhood to obtain the corrected input, and binarized with a threshold of 0.5; subsequently, the quantization error is filtered by weights. Assign to neighboring pixels.

[0026] for In the phase-shift method, the captured fringe pattern can be expressed by the following formula: , in , , Representing pixel coordinates Background light intensity, stripe contrast, and phase at the location.

[0027] , The calculation result only represents the wrapped phase; the final absolute phase value still needs to be obtained through the phase unrolling step.

[0028] Step 2: With the projector focused, the binary stripe pattern is sequentially projected onto the surface of the object being measured; and the deformed stripe image sequence modulated by the surface of the object being measured is acquired; The projection defocusing process is usually modeled in simulation as A Gaussian filter, whose standard deviation is set to... ,in , , These correspond to weak, medium, and strong defocus levels, respectively. In contrast, this embodiment completely eliminates the defocusing step and maintains focused projection throughout the entire 3D measurement process. However, even under focused conditions, the binary pattern projected onto the object surface does not present a sharp binary halftone image composed of black and white blocks; instead, each optomechanical pixel projected onto the object surface exhibits a continuous intensity distribution with the brightest center and gradually decreasing towards the edges. This phenomenon originates from the point spread function (PSF) of the projector's optical system, which describes the inherent optical blurring characteristics of the projector.

[0029] Step 3: Extract phase information based on the deformed stripe image sequence and perform three-dimensional morphology reconstruction.

[0030] To verify the feasibility of the proposed method in 3D topography measurement, a structured light experimental system was built, including a DLP projector (TI DLPLCR4500EVM) and an industrial camera (Alvium 1800 U-2050m). The projector and camera have resolutions of 912×1140 and 5496×3672, respectively. After system calibration and establishing a complete mapping from phase to 3D coordinates, an ideal 8-bit sinusoidal fringe under focused projection was first used to scan a limestone face statue, and the reconstructed result was used as a reference for the true value. Subsequently, using the binary fringe pattern generated by the proposed method, and the binary fringe pattern generated by the improved error diffusion algorithm, 3D topography measurements were performed on the statue under different focusing states (including focused, weakly defocused, moderately defocused, and strongly defocused). The fringe width was 10 pixels, and the phase shift step was 4.

[0031] The final 3D reconstruction result is as follows Figure 2 As shown. To more clearly compare the reconstruction differences, this embodiment extracted the depth data of the 3800th column in each experiment and plotted its corresponding contour error distribution map. Figure 3 This shows the contour comparison of column 3800 in each group of experiments. From Figure 2 and Figure 3 It can be seen that, regardless of whether the object surface contour is gentle or steep, the height error of the method described in this embodiment is lower than that of the benchmark method under all defocus conditions, with a smaller standard deviation, demonstrating stronger robustness. Furthermore, both the method described in this embodiment and the benchmark method achieve optimal results under focused projection conditions. Further observation reveals that excessive defocus significantly reduces fringe contrast, leading to a substantial increase in measurement error, which is one of the inherent bottlenecks of traditional binary defocusing techniques. Table 1 summarizes the phase and height errors of eight sets of measurement results. The method proposed in this embodiment demonstrates superior measurement accuracy under focused conditions compared to all comparative methods, and is superior to the benchmark scheme under different degrees of defocus. Experimental results show that the binary projection contouring method proposed in this embodiment achieves 3D reconstruction accuracy comparable to existing binary defocusing techniques without requiring manual projection defocusing. It effectively simplifies the measurement process, saves focusing time and costs, improves the effective measurement depth of field, and enhances generalization capabilities across different hardware platforms.

[0032] Table 1. Root Mean Square Error of Three-Dimensional Measurement Experiment Results

[0033] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A binary fringe projection profilometry method based on improved error diffusion, characterized in that, Includes the following steps: S1. Obtain multiple ideal stripe images with sinusoidal grayscale distribution, and generate corresponding 1-bit binary stripe patterns based on the improved error diffusion algorithm; The improved error diffusion algorithm includes using the Sierra-Lite error diffusion weight matrix for error allocation; introducing random perturbations during weight allocation; and traversing image pixels in a serpentine scanning order. S2. With the projector focused, the binary stripe pattern is sequentially projected onto the surface of the object being measured; and the deformed stripe image sequence modulated by the surface of the object being measured is acquired. S3. Extract phase information based on the deformed stripe image sequence and perform three-dimensional morphology reconstruction.

2. The binary fringe projection profilometry method based on improved error diffusion according to claim 1, characterized in that, The random disturbances are grouped according to their weights, and the amplitude of the random disturbance in each group does not exceed 50% of the smallest weight in that group.

3. The binary fringe projection profilometry method based on improved error diffusion according to claim 1, characterized in that, The serpentine scanning sequence is to scan even-numbered rows of the image from left to right and odd-numbered rows from right to left.

4. The binary fringe projection profilometry method based on improved error diffusion according to claim 1, characterized in that, The "focus state" refers to the projector's optical system being in the state of optimal image clarity, without the need for additional defocusing operations.

5. The binary fringe projection profilometry method based on improved error diffusion according to claim 1, characterized in that, Phase information is extracted based on the deformed stripe image sequence, and three-dimensional topography reconstruction is performed, including: The wrapping phase is calculated from the deformed stripe image sequence based on the multi-step phase shift method; The absolute phase distribution is obtained by performing phase unrolling on the wrapped phase. Based on the system calibration parameters, the absolute phase is mapped to three-dimensional spatial coordinates.

6. The binary fringe projection profilometry method based on improved error diffusion according to claim 5, characterized in that, The multi-step phase-shifting method has no fewer than 3 phase-shifting steps.

7. The binary fringe projection profilometry method based on improved error diffusion according to claim 1, characterized in that, The projection frame rate of the binary stripe pattern is not less than 1000Hz, and it has a quasi-sinusoidal light intensity distribution under focused projection.

Images