A non-uniform line width encoding scanning reconstruction method for high dynamic range objects

Abstract

This invention provides a non-uniform linewidth coded scanning reconstruction method for high dynamic range objects. The method mainly includes the following steps: 1) Generating a non-uniform linewidth line scan sequence pattern and a white light pattern, and projecting them onto the surface of the high dynamic range object using an optical encoding device, while an image acquisition device simultaneously acquires the deformed image; 2) Calibrating the surface reflection feature values ​​using the reflection feature value calibration method proposed in this invention, and then equalizing the line scan image based on the reflection feature values; 3) Binarizing the equalized line scan image using a binarization method based on clustering to obtain a standard binary image; 4) Using sinusoidal weighting coefficients to... n Zhang's binary image is synthesized as m 5) Extract the wrapping phase from the sinusoidal image and identify the fringe order to obtain the absolute phase; 6) Import the absolute phase into the pre-calibrated system model to complete the final three-dimensional measurement. The non-uniform linewidth encoding scanning reconstruction method for high dynamic range objects provided by this invention only requires the non-uniform linewidth encoding strategy, surface reflection feature value calibration method, and sinusoidal synthesis model provided by this invention to eliminate image overexposure and achieve high-precision measurement of high dynamic range objects.

Description

Technical Field

[0001] This invention relates to a three-dimensional reconstruction method, and more specifically, to a non-uniform linewidth encoded scanning reconstruction method for objects with high dynamic range. Background Technology

[0002] With significant advancements in intelligent manufacturing and measurement technologies, 3D topography measurement has found widespread application in various industries, including aerospace, industrial inspection, biomedicine, and gaming. Typical applications include analyzing gear operation, identifying surface defects in ship propellers, and inspecting the forming accuracy of streamlined aircraft surfaces. Fringe projection profilometry, as a typical 3D topography measurement technique, possesses several significant advantages, such as the ability to acquire high-density point cloud data, high measurement accuracy, and relatively low system setup costs. In industrial measurement, objects with high or multiple reflectivities are defined as high dynamic range (HDR) objects, and the common method for measuring the surfaces of such objects is called high dynamic range technology. However, due to the large range of reflectivity variations on the object's surface, coupled with the limited grayscale response range of cameras, using fringe projection profilometry for 3D measurement of these objects can easily lead to overexposure or underexposure of the image, resulting in a significant decrease in measurement accuracy. Accurate 3D topographic measurement of HDR objects remains a major unresolved technical challenge in the field of 3D measurement. The reflective properties of high dynamic range (HDR) objects can easily cause the intensity of reflected light to exceed the dynamic range of the camera sensor. Since the spectral information of the object's surface cannot be fully recorded in a single shot, the image may exhibit both overexposure and underexposure. Both overexposure and underexposure distort the fringe pattern information, severely impacting subsequent phase extraction and unfolding processes, leading to phase calculation errors and significant 3D errors in the reconstructed model, greatly reducing measurement accuracy. Currently, the severe decrease in measurement accuracy when dealing with HDR objects has limited the practical applicability of fringe projection profilometry. Summary of the Invention

[0003] The purpose of this invention is to provide a non-uniform linewidth coded scanning reconstruction method for high dynamic range (HDR) objects. The main content of this method is to design a line scan pattern and project the designed line scan pattern and a white light pattern onto the surface of the HDR object using an optical encoding device, while an image acquisition device simultaneously acquires the image. The surface reflection feature value calibration method proposed in this invention is used to calibrate the reflection feature values ​​of the object's surface. Based on the reflection feature value equalization, the line scan image can be efficiently binarized. Then, the sinusoidal image synthesis model proposed in this invention is used to complete the conversion from a binary image to a sinusoidal image. Next, the fringe order of the enclosing phase and the identification phase contained in the sinusoidal image is extracted to obtain a continuous absolute phase. Finally, the absolute phase is imported into a pre-calibrated system model to complete the final three-dimensional measurement. This non-uniform linewidth coded scanning reconstruction method for HDR objects utilizes the characteristic that binary patterns are insensitive to surface reflectivity, which can simultaneously improve the measurement accuracy and efficiency of HDR objects. To achieve the above objectives, the technical solution provided by this invention is: a non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range, the method specifically including the following... Figure 1 All the steps described: Step 1: Generate a pattern sequence, as shown in the attached image. Figure 2 Generate a white light pattern for calibrating surface reflection characteristics. It has the following characteristics: the grayscale value of all its pixels is 1 (normalized grayscale value). A non-uniform linewidth scan sequence pattern is generated for recovering phase information, and it has the following characteristics: all patterns encode grayscale only in a single direction, that is, grayscale only changes in the horizontal or vertical direction; each scan line, starting from the initial pattern, is shifted by a single linewidth pixel distance after each projection, forming a continuous scan until all pixels are covered, resulting in... N The scan pattern has two widths: 1 pixel and 2 pixels. The area covered by each scan line in the entire pattern sequence is called a period, where the scan line corresponding to a 1-pixel line width is called a narrow period (including...). N (2 pixels) and marked as 0, the scan line corresponding to a line width of 2 pixels is called the wide period (containing 2 pixels). N Each pixel is marked as 1; the width and narrow periods are encoded according to the binary cyclic sequence B{2,5} (000001000110010…), forming a sequence containing 5 pixels. A subsequence (00000, 00001, 00010, ...). Convert the subsequence to decimal and create a stripe level lookup table. Step 2: Obtain the deformed image sequence, as shown in the attached image. Figure 3 The generated line scan sequence pattern and white light pattern The input is an optical encoding device, which sequentially projects all patterns onto the surface of the object to be measured, and an image acquisition device simultaneously acquires the images. The acquired line scan images are denoted as... The white light image is denoted as , Among them, controlling the white light image The exposure does not result in overexposure, and the projection is repeated. G Second-rate( G >10). Step 3: Calibrate the reflection characteristic values, as shown in the attached figure. Figure 4 The white light images are sequentially superimposed according to the order of projection acquisition. The superposition process is as follows: After stacking, G units will be obtained. Value. Based on The grayscale growth rate is calculated using the following procedure. : Will As a characteristic value of surface reflection. Step 4: Equalize grayscale distribution. Based on surface reflection characteristic values. R The process of equalizing the line scan image is as follows: A balanced line scan image. Step 5: Image Binarization. Perform binarization and clustering on the balanced line scan image, and set the initial cluster centers. and And divide all pixels into 2 Cluster and Cluster center optimization is performed as follows: An iterative process is introduced to optimize binary classification. In each iteration, pixels are assigned to each cluster by judging the difference between pixel grayscale and cluster center, and the assignment formula is as follows: in, f This represents the number of iterations. The average value of each cluster is selected as the new cluster center for the next calculation, as shown below: in and They represent the first f The number of pixels in the two clusters during the next iteration. When the difference... When the value is less than a preset threshold, the iteration stops and the current value is used as the optimized cluster center. Finally, based on... and Binarization is performed according to the following formula. in, Indicates the first n Zhang standard binary image. Step 6: Synthesize the sine wave image, as shown in the attached image. Figure 5 Using sinusoidal weighting coefficients, n Zhang's binary image Synthesized m Sine images with different phase shifts The process is as follows: in, It is a synthesized sine wave image. It is the phase shift step number, which includes the phase shift amount. . Step 7: Identify the stripe levels, as shown in the attached diagram. Figure 6 Based on sinusoidal image sequences The phase shift algorithm is used to calculate the wrapping phase. The phase is identified using the following formula. Feature points: Using two adjacent feature points as period endpoints, the pixel length of each period is calculated, with narrow periods marked as 0 and wide periods marked as 1. A sliding window with a width of 5 is set to span five periods, and the width-to-width composition order of these five periods, A1A2A3A4A5 (e.g., 00001), is identified. This order is then converted to decimal using the following formula: Based on the lookup table established using the cyclic sequence in step one, the fringe order of the first period within the window can be determined. By iterating through the entire wrapped phase using the above operations, the fringe order map can be obtained. . Step 8: Combine the fringe order and the wrapping phase to obtain the unfolded absolute phase. The process is as follows: absolute phase Import the pre-calibrated system model to complete the final 3D measurement. The beneficial effects of the non-uniform linewidth coded scanning reconstruction method for high dynamic range objects of the present invention are: ① This non-uniform linewidth coding scanning reconstruction method for high dynamic range objects addresses the problem of severe accuracy degradation when measuring high dynamic range objects using stripe projection profilometry. It proposes a non-uniform linewidth coding strategy to suppress image overexposure while avoiding additional projection patterns. Furthermore, by combining a pixel-level reflection feature value calibration method and a sinusoidal synthesis model, high-precision 3D reconstruction of high dynamic range objects can be achieved. ② The proposed non-uniform linewidth encoding strategy introduces a binary cyclic sequence that can encode the level information into the line scan pattern through the period width, avoiding the need to project additional patterns for phase unfolding, thereby improving the reconstruction efficiency; ③ The proposed reflection feature value calibration method can perform pixel-level calibration of the reflection feature value of a high dynamic object surface, and equalize the line scan pattern based on the reflection feature value, thereby achieving robust binarization processing. Theoretically, this processing can eliminate signal weaknesses including overexposure and random noise. ④ The derived sinusoidal synthesis model can convert standard binary images into high-quality, high-depth sinusoidal fringe images, thereby providing image support for subsequent phase extraction; Attached Figure Description Figure 1 This is a schematic diagram of the overall process of the non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range according to the present invention. Figure 2 This is a schematic diagram of the line scan pattern encoding strategy of the non-uniform linewidth encoding scan reconstruction method for objects with high dynamic range according to the present invention. Figure 3 This is a schematic diagram of the line scanning process of the non-uniform linewidth encoded scanning reconstruction method for objects with high dynamic range according to the present invention. Figure 4 This is a schematic diagram of the surface reflection feature value calibration process of the non-uniform linewidth encoded scanning reconstruction method for objects with high dynamic range according to the present invention. Figure 5 This is a schematic diagram of the sinusoidal image synthesis process of the non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range according to the present invention. Figure 6 This is a schematic diagram of stripe level recognition in the non-uniform linewidth coding scanning reconstruction method for objects with high dynamic range according to the present invention. Figure 7 This is a schematic diagram of the measurement system for the non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range according to the present invention. Detailed Implementation The following description is intended to disclose the present invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art. The basic principles of the invention defined in the following description can be applied to other embodiments, modifications, improvements, equivalents, and other technical solutions that do not depart from the spirit and scope of the invention. Referring to the above figures, this invention discloses a non-uniform linewidth coded scanning reconstruction method for high dynamic range objects, which is implemented through a projector-camera measurement device. This measurement device mainly consists of a computer 1, a camera 2, and a projector 3. The computer 1 is connected to both the industrial camera 2 and the projector 3, allowing the computer 1 to control the projector 3 to project a stripe pattern onto the surface of the object being measured, while simultaneously controlling the industrial camera 2 to synchronously acquire images and send them to the computer 1 for subsequent phase calculation and 3D reconstruction. The high dynamic range object 4 is located in front of the camera 2 and the projector 3, and this arrangement should ensure that the range of the high dynamic range object is covered by the common projection area of ​​the projector 3 and the camera 2. Referring to the accompanying drawings, this invention discloses a non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range. The specific implementation steps are as follows: Step 1: Set up the measurement system, as shown in the attached document. Figure 7 A high dynamic range object 4 is located in front of camera 2 and projector 3, with camera 2 and projector 3 positioned to the left and right of the high dynamic range object 4, respectively, and the range of the high dynamic range object being within the common field of view of both. Computer 1 is electrically connected to camera 2 and projector 3 respectively, and an intrinsic and extrinsic parameter model of the projector-camera system is established through calibration. Step 2: Computer 1 generates a line scan pattern (one cycle contains...) N =16 pixels) and a white light pattern. Generate the corresponding pixel length based on the projector resolution. H= 1140 and corresponding pixel width W= 912 16 line scan patterns; select the number of periods for the entire stripe pattern according to measurement needs. T =57. Step 3: The computer 1 projects the pattern generated in Step 2 onto the surface of the high dynamic range object sequentially through the projector 3, obtaining... The white light pattern is repeatedly projected. G =20 times, and at the same time, camera 2 synchronously acquires the modulated deformed image, and the acquired image is sent to computer 1. Step 4: The distorted white light image acquired in Step 3 The time series data are overlaid using the following formula: Calculate according to the following formula Follow g growth rate R , Step 5: Use the growth rate obtained in Step 4 R The deformed line scan image obtained in step 3 Perform equalization to obtain the equalized line scan image. The equalization formula is as follows: Step 6: Binarize the balanced line scan image obtained in Step 5 based on the clustering concept. Set the initial cluster centers. This will divide all pixels into 2 clusters. and An iterative process is introduced to optimize binary classification. In each iteration, pixels are assigned to each cluster by judging the difference between pixel grayscale and cluster center, using the following allocation formula: Step 7: Based on the two clusters obtained in Step 6, select the average value of each cluster as the new cluster center for the next calculation, as shown below: When the difference When the value is less than a preset threshold (set to 0.1), the algorithm converges and the iteration terminates. Finally, based on... and The pixels are divided into two clusters using the nearest neighbor classification principle, and binarization is performed according to the following formula. Step 8: Using sinusoidal weighting coefficients, transform the binary image obtained in Step 7... Synthesized into a sine wave image The synthesis formula is as follows: Step 9: Extract the wrapping phase contained in the sinusoidal image and identify the fringe order. The wrapping phase is calculated using a traditional phase-shifting algorithm. The phase is identified as The feature points are used, with adjacent feature points as period endpoints, and the pixel length of each period is calculated, where narrow periods are marked as 0 and wide periods as 1. A sliding window with a width of 5 is set to span five periods, and the order of the width and narrowness of these five periods is identified. Taking the sequence 00100 as an example, this sequence is converted to the decimal number 4. Then, based on the level lookup table built on the cyclic sequence, the stripe level of the first period within the window can be determined. By traversing the above operations throughout the entire wrapped phase, a stripe level map can be formed. . Step 10: By combining the fringe order and the wrapping phase, the unfolded absolute phase can be obtained and imported into the pre-calibrated system model to complete the final three-dimensional measurement. The above embodiments are only used to illustrate the present invention. The parameters involved in each step, as well as the structure, size, placement, and shape of each component, can vary. Any improvements or equivalent transformations made to individual components based on the principles of the present invention should not be excluded from the protection scope of the present invention. The above provides a detailed description of one embodiment of the present invention, but the content is only a preferred embodiment and should not be considered as limiting the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range, characterized in that, This method mainly includes the following steps: Step 1: Generate a white light pattern for calibrating surface reflection characteristics. It has the following characteristics: all its pixels have a grayscale value of 1 (normalized grayscale value); it generates a non-uniform linewidth line scan sequence pattern for recovering phase information. It has the following characteristics: all patterns encode grayscale only in a single direction, meaning grayscale only changes in the horizontal or vertical direction; each scan line, starting from the initial pattern, shifts by a single linewidth pixel after each projection, forming a continuous scan until all pixels are covered, resulting in... N The scan pattern has two widths, containing 1 pixel and 2 pixels respectively. The coverage area of ​​each scan line in the entire pattern sequence is called a period, where the scan line corresponding to a 1-pixel line width is called a narrow period (including...). N (2 pixels) and marked as 0, the scan line corresponding to a line width of 2 pixels is called the wide period (containing 2 pixels). N Each pixel is labeled as 1; the width and narrow periods are encoded according to the binary cyclic sequence B{2,5} (000001000110010…), forming a 2-bit sequence of length 5. 5 Each subsequence (00000, 00001, 00010, ...) is converted into decimal and recorded in a stripe level lookup table. Step 2: Generate the line scan sequence pattern and white light pattern The input is an optical encoding device, which sequentially projects all patterns onto the surface of the object to be measured, and an image acquisition device simultaneously acquires the images. The acquired line scan images are recorded as follows: The white light image is denoted as , Among them, controlling the white light image The exposure does not result in overexposure, and the projection is repeated. G Second-rate( G >10); Step 3: Superimpose the white light images sequentially according to the projection order. The superposition process is as follows: After stacking, G units will be obtained. Value; based on The grayscale growth rate is calculated using the following procedure. : Will As a characteristic value of surface reflection; Step 4: Based on surface reflection characteristic values R The process of equalizing the line scan image is as follows: in, For a balanced line scan image; Step 5: Perform binarization and clustering on the balanced line scan image, using the initial cluster centers. and Divide all pixels into 2 Cluster and Cluster center optimization is performed as follows: An iterative process is introduced to optimize binary classification. In each iteration, pixels are assigned to each cluster by judging the difference between pixel grayscale and cluster center, and the assignment formula is as follows: in, f This represents the number of iterations. The average value of each cluster is selected as the new cluster center for the next calculation, as shown below: in and They represent the first f The number of pixels in the two clusters during the next iteration. When the difference... When the value is less than a preset threshold, the iteration stops and the current value is used as the optimized cluster center; finally, based on... and And perform binarization according to the following formula. in, Indicates the first n Zhang's standard binary image; Step 6: Use sine weighting coefficients to... n Zhang's binary image Synthesized m Sine images with different phase shifts The process is as follows: in, It is a synthesized sine wave image. It is the phase shift step number, which includes the phase shift amount. . Step 7: Based on the sinusoidal image sequence The phase shift algorithm is used to calculate the wrapping phase. The phase is identified using the following formula: Feature points: Using two adjacent feature points as period endpoints, the pixel length of each period is calculated, with narrow periods marked as 0 and wide periods marked as 1. A sliding window with a width of 5 is set to span five periods, and the width-to-short composition order of these five periods, A1A2A3A4A5, is identified. This order is then converted to decimal using the following formula: Based on the lookup table established using the cyclic sequence in step 1, the fringe order of the first period within the window can be determined; by traversing the entire wrapped phase using the above operations, the fringe order map can be obtained. ; Step 8: Combining the fringe order and the wrapping phase yields the unfolded absolute phase, as follows: absolute phase Import the pre-calibrated system model to complete the final 3D measurement.

2. The non-uniform linewidth coded scanning reconstruction method for objects with high dynamic range as described in claim 1, characterized in that, The binary cyclic sequence B{2,5} (000001000110010…) introduced in step 1 can be rewritten as B{2,t} (…). …), the characteristic of this binary cycle is that it contains 2 t Subsequences ( , , ...).

Images