A projection parallel single-pixel measurement method based on time multiplexing coding

Abstract

The application discloses a kind of projection parallel single-pixel measurement methods based on time multiplexing coding, first using projector to the object to be measured cyclic projection is constituted by full white pattern, full black pattern and four step phase shift base stripe graph sequence graph, camera synchronously shoots the object to be measured, then on the projection sequence similar full white pattern and full black pattern, calculate background light intensity, calculate spectral coefficient, finally all camera shooting image pixel executes Fourier single-pixel imaging in parallel.This application can reduce the number of projection patterns by time multiplexing parallel single-pixel coding method, while ensuring the effectiveness of parallel single-pixel measurement, greatly improve the coding efficiency.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of computer vision three-dimensional measurement calculation, and particularly relates to a projection parallel single-pixel measurement method based on time multiplexing coding. BACKGROUND

[0002] In recent years, parallel single-pixel measurement technology has high adaptability and high signal-to-noise ratio, and can realize high-adaptive three-dimensional reconstruction for complex light and reflection scenes such as diffuse reflection, specular reflection, dark area, subsurface scattering and interreflection, and has been increasingly widely applied in the field of computer vision measurement calculation.

[0003] With the rapid development of computer technology and the continuous updating of projection and imaging devices, parallel single-pixel measurement technology is extended to high-speed dynamic three-dimensional scenes, but the existing method still needs to project hundreds of base patterns to complete a measurement on the object to be measured, and the reconstruction efficiency is low. SUMMARY

[0004] The present application aims to overcome the shortcomings of the prior art, and provides a projection parallel single-pixel measurement method based on time multiplexing coding.

[0005] In order to achieve the above-mentioned purpose, the technical scheme adopted by the present application to solve its technical problems is as follows: a projection parallel single-pixel measurement method based on time multiplexing coding, comprising the following steps:

[0006] S1, using a projector to cyclically project a sequence of patterns composed of a full white pattern I W , a full black pattern I B and a four-step phase shift base fringe pattern to an object to be measured, and a camera synchronously photographs the pattern of the object to be measured;

[0007] S2, calculating the background light intensity of the similar full white pattern I W and the full black pattern I B in the projection sequence;

[0008] S3, assisting in calculating the spectral coefficients according to the calculated background light intensity;

[0009] S4, performing Fourier single-pixel imaging on all the image pixels (u c , v c ) photographed by the camera in parallel.

[0010] Further, the four-step phase shift base fringe pattern in step S1 is generated by the following formula:

[0011]

[0012] The stripe frequency and direction of the phase-shifted stripes in the four-step phase-shifted base fringe pattern are different, and correspond to different frequency spectrum coefficients in Fourier single-pixel imaging The sequence S of the four-step phase-shifted base fringe pattern is cyclically arranged according to the order of I1, I2 or I3, I4 and the frequency spectrum coefficients

[0013]

[0014] The above two stripe combinations of I1, I2 or I3, I4 can be selected at will, and the sequence S corresponds to the same frequency spectrum and N different coefficients Next, the all-white pattern I W and the all-black pattern I B are sequentially inserted at the beginning and the end of the sequence S to form a cyclic sequence: I W S1I B S2I W S3I B ... , and the sequence pattern is cyclically projected to the object to be measured.

[0015] Further, the step S2 comprises: in the collected image sequence, each combination I W S i I B or I B S i I W A three-dimensional result can be calculated; wherein the frequency spectrum coefficients can be calculated by using S i The left and right adjacent white patterns and black patterns are averaged to calculate the extracted background light intensity For the frequency spectrum coefficient numbered n, the background light intensity A(u c ,v c ,i,n) is calculated, and then the two fringe patterns with the same frequency in the unit are used to calculate the frequency spectrum coefficient

[0016]

[0017] The frequency spectrum coefficients are sequentially calculated according to the sampling path and filled into the frequency spectrum, and finally the reconstructed frequency spectrum is inversely Fourier transformed to form an image, and the intensity distribution of the image reflects the energy corresponding relationship between the image pixels (u c ,v c ) and the entire projector plane, that is, the energy of the current point comes from which positions on the target surface of the projector.

[0018] The beneficial effects of the present application are: the present application uses a time-multiplexed parallel single-pixel encoding method, which can reduce the number of projection patterns and greatly improve the encoding efficiency while ensuring the effectiveness of parallel single-pixel measurement. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1 This is a flowchart of a traditional parallel single-pixel measurement method.

[0020] Figure 2 This is a flowchart of the parallel single-pixel measurement method based on time multiplexing according to the present invention;

[0021] Figure 3 This is a schematic diagram of the projection pattern sequence of the parallel single-pixel measurement method based on time multiplexing according to the present invention. Detailed Implementation

[0022] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0023] The general workflow for parallel single-pixel 3D measurement is as follows: Figure 1 As shown:

[0024] S1, using a projector, cyclically project a pure white pattern I onto the object to be measured. W All-black pattern I B The sequence of images, consisting of a four-step phase-shifted base fringe pattern, is simultaneously captured by the camera to measure the pattern of the object under test;

[0025] S2 uses Fourier single-pixel imaging for all image pixels captured by the camera;

[0026] S3, use epipolar constraints and grayscale value centroid method to extract the pixel coordinates of the direct illumination component on each imaging plane;

[0027] S4 converts pixel coordinates into point clouds through system calibration.

[0028] The above measurement method first projects a series of Fourier basis fringe patterns, and then captures the projected image with a camera at a certain angle to the projection direction; each pixel on the camera target surface is regarded as a single-pixel detector, and as the projection basis pattern changes, the single pixel (u) on the camera plane... c ,v c The camera will receive the corresponding changes in light intensity. Further, following the Fourier single-pixel method, two-dimensional single-pixel imaging is performed on each pixel on the camera target surface. The intensity distribution of the resulting image reflects the camera pixel (u... c ,v c The energy correspondence between the current point (u) and the entire projector plane, i.e., the energy correspondence between the current point (u) and the entire projector plane. c ,v c The energy received comes from which parts of the projector target surface?

[0029] By further applying epipolar constraints to the imaging plane, the pixel (u) can be determined. c ,v c Received illumination components: Select the direct illumination component (u) for 3D reconstruction.d ,v d ), and the indirect lighting component that causes interference (u id ,v id Using the gray-scale centroid method, (u) can be further determined. d ,v d The pixel position of the camera pixel is then determined through system-calibrated transformation. c ,v c Projector plane point (u) d ,v d The correspondence between world coordinates (X, Y, Z) is established. By performing the above single-pixel imaging, direct illumination component positioning, and coordinate transformation in parallel on all pixels on the camera target surface, a complete 3D point cloud of the object can be obtained.

[0030] Reference Figure 2 As shown, the present invention discloses a projection parallel single-pixel measurement method based on time multiplexing coding, which includes the following steps.

[0031] S1, using a projector, cyclically project a pure white pattern I onto the object to be measured. W All-black pattern I B The sequence of images, consisting of a four-step phase-shifted base fringe pattern, is captured by the camera simultaneously to obtain the pattern of the object to be measured.

[0032] This step uses the following formula to generate a four-step phase-shifted substrate fringe pattern.

[0033]

[0034] The fringe frequencies and orientations of the phase-shifted fringes in the four-step phase-shifted substrate fringe pattern are different, corresponding to different spectral coefficients in Fourier single-pixel imaging. The sequence S of the four-step phase-shifted substrate fringe pattern is cyclically arranged according to the order of I1, I2 or I3, I4 and the spectral coefficients:

[0035]

[0036] Either the above combination of stripes I1, I2 or I3, I4 can be chosen, and sequence S corresponds to N different coefficients under the same spectrum. Next, arrange the all-white pattern I in sequence. W and all-black pattern I B Inserting into the beginning and end of sequence S forms a cyclic sequence: I W S1I B S2I W S3I B ...and cyclically project the sequence diagram onto the object to be measured.

[0037] S2, for similar all-white patterns I in the projection sequence. Wand all-black pattern I B Calculate the background light intensity; for example, black s white s black s white, calculate the background light intensity for the first black and the first white, but not for the second white.

[0038] This step involves acquiring the image sequence, and for each combination I... W S i I B or I B S i I W A three-dimensional result can be calculated; S can be used when calculating the spectral coefficients. i The background light intensity is extracted by averaging the values ​​of adjacent white and black patterns. For the spectral coefficient numbered n, calculate the background light intensity A(u c ,v c After i, n), use the unit with the same frequency. Two stripe patterns are sufficient to calculate the spectral coefficients.

[0039]

[0040] The spectral coefficients are calculated sequentially according to the sampling path and filled into the spectrum. Finally, an inverse Fourier transform is performed on the reconstructed spectrum for imaging. The intensity distribution of the resulting image reflects the pixel (u) of the image. c ,v c The energy correspondence between the current point and the entire projector plane, that is, from which positions on the projector target surface the energy received at the current point comes.

[0041] S3, based on the calculated background light intensity, assists in calculating the spectral coefficients.

[0042] S4, for all camera image pixels (u c ,v c Fourier single-pixel imaging is performed in parallel.

[0043] Compared to existing parallel single-pixel 3D measurement methods and the current three-step phase-shift coding method with the fewest patterns, the measurement method of this invention, while retaining measurement effectiveness, only uses two fringe patterns to calculate the spectral coefficient of a frequency, greatly compressing the redundancy in the coding (retaining only 2 / 3 of the patterns of the original method), and improving measurement efficiency by 33%. A schematic diagram of the projected pattern sequence is shown below. Figure 3 As shown.

[0044] The above embodiments are merely illustrative of the principles and effects of the present invention, as well as some of the application examples. For those skilled in the art, various modifications and improvements can be made without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A time-multiplexed encoding based projection parallel single-pixel measurement method, characterized in that: The method comprises the following steps S1, using a projector to project a sequence pattern composed of full white pattern I W , full black pattern I B and four-step phase shift base fringe pattern to the object to be measured, and a camera synchronously captures the object to be measured; wherein the formula for generating the four-step phase shift base fringe pattern is: , The stripe frequencies and directions of the phase-shifted stripes in the four-step phase-shifted base fringe pattern are different, and correspond to different spectral coefficients in Fourier single-pixel imaging , in the order of I 2 ,I 1 or I 3 ,I The sequence S of the four-step phase-shifted base fringe pattern is arranged in the order of the spectral coefficients and circularly ， The above I 2 ,I 1 or I 3 ,I 4 either one of two stripe combinations, in sequence S corresponding to the same spectrum, N different coefficients ; in turn, a full white pattern I W and a full black pattern I B insertion sequence S head and tail, forming a cyclic sequence: and cyclically projecting the sequence pattern to the object to be measured; S2, for the projection sequence on adjacent full white pattern I W and full black pattern I B Calculate background light intensity: in the collected image sequence, each combination I W S i I B or I B S i I W Calculate a three-dimensional result; when calculating the spectral coefficients, use S i Left and right adjacent white pattern and black pattern, take the average to calculate the extracted background light intensity A u c , v c , i , n ):​ ， For spectral coefficients with number n the background light intensity A ( u c , v c , i , n ) is calculated and then the spectral coefficients are calculated using two fringe patterns with the same frequency : ， Spectrum coefficients are calculated in sequence according to a sampling path and filled into a spectrum, and finally inverse Fourier transform is performed on the reconstructed spectrum to image; S3, calculating the spectrum coefficients according to the calculated background light intensity; S4, calculating the spectrum coefficients according to the calculated background light intensity. S4, for all camera image pixels ( u c ,v c Fourier single-pixel imaging is performed in parallel.

Images