Improved phase-height mapping method in three-dimensional vision measurement based on hybrid device pose

Abstract

The application discloses an improved phase-height mapping method in three-dimensional vision measurement based on a mixed device posture, and the principle is as follows: the improved phase-height mapping method in three-dimensional vision measurement based on a mixed device posture, first, a projector projects a stripe image to a measurement target surface, then, an unwrapped phase is obtained by analyzing a stripe image captured by a camera, the unwrapped phase is a difference value between a measurement target unwrapped phase and a measurement target unwrapped phase, then, a GRNN model is used to weaken a phase information offset at the same height caused by device posture randomness, finally, the unwrapped phase after the offset is eliminated is substituted into a phase-height mapping formula to complete mapping of the phase information to the height information. The improved phase-height mapping method in three-dimensional vision measurement based on a mixed device posture can realize high-quality phase-height mapping only by using a small amount of phase-height mapping formulas, and the practicability of the phase-height mapping is improved.

Description

Technical Field

[0001] This invention belongs to the field of non-contact 3D vision measurement technology, and relates to an improved phase-height mapping method in 3D vision measurement based on hybrid device posture. Background Technology

[0002] With the continuous development of artificial intelligence technology, intelligent manufacturing is an inevitable direction for the development of the manufacturing industry. In most industrial production processes, the three-dimensional information of objects is one of the important pieces of information for realizing intelligent manufacturing. Therefore, how to achieve non-contact, high-quality three-dimensional measurement of objects is a key technology in the field of intelligent manufacturing. The phase measurement profilometry (PMP) system, composed of a projector and a camera, has been widely used in industrial production due to its advantages such as simple structure, strong anti-noise interference capability, and full-field non-contact measurement. The principle of this system to achieve three-dimensional measurement is as follows: the projector first projects a fringe image onto the target to be measured, then analyzes the fringe image captured within the field of view of the camera to obtain the unfolded phase, and then uses a calibrated phase-height mapping model to map the unfolded phase containing the height information of the target to be measured into the actual height value. The phase-height mapping is a crucial part of this process, so how to achieve accurate phase-height mapping simply and effectively is one of the research goals in this direction. However, existing phase-height explicit mapping methods and phase-height implicit mapping methods both have some problems. The phase-height explicit mapping method requires the calibration of many system parameters, which not only makes the calibration process complex but also easily introduces additional errors. The phase-height implicit mapping method is currently more commonly used. Although this method does not require a cumbersome calibration process for system parameters and can obtain the implicit parameters from the phase information corresponding to several known heights, thus obtaining the mapping relationship between phase and height, in practice, the phase measurement profilometry system is easily affected by the device's posture, causing the phase information to shift. Furthermore, this method requires a mapping formula for each pixel position, thus requiring a large amount of memory space to store the above mapping formula, which greatly limits the application of this method. Summary of the Invention

[0003] The purpose of this invention is to provide an improved phase-height mapping method for 3D vision measurement based on hybrid device posture, which solves the problems that the accuracy of existing implicit phase-height mapping methods is easily affected by device posture factors, or that a large number of phase-height mapping formulas are required to achieve high-quality phase-height mapping.

[0004] To achieve the above objectives, the technical solution adopted in this invention is as follows: an improved phase-height mapping method in 3D visual measurement based on mixed device posture. First, a stripe image is projected onto the surface of the target to be measured using a projector. Then, the unfolded phase obtained by analyzing the stripe image captured by the camera is analyzed. The unfolded phase is the difference between the unfolded phase with and without the target to be measured. Then, a GRNN model is used to reduce the phase information offset caused by the randomness of the device posture at the same height. Finally, the unfolded phase with the offset eliminated is substituted into the phase-height mapping formula to complete the mapping from phase information to height information.

[0005] As a preferred technical solution of the present invention, it is implemented according to the following steps:

[0006] Step 1: Using formula (1), the coordinates (x, y) of the unfolded phase shift of the stripe image are taken as input, and the values ​​of the unfolded phase shift at the corresponding coordinates are taken as output, to train the GRNN model.

[0007] ψ offset =GRNN(x,y) (1)

[0008] In the formula: ψ offset To expand the phase offset at the corresponding coordinates;

[0009] Step 2, using formula (2), let ψ minus ψ offset The expanded phase, denoted as ψ′, is obtained after eliminating the effect of the expanded phase shift.

[0010] ψ′=ψ-ψ offset (2)

[0011] In the formula: ψ is the expanded phase;

[0012] Step 3: Substitute ψ′ as input into formula (3) to obtain a high-quality phase-height mapping result;

[0013]

[0014] In the formula: h(x,y) is the actual height value corresponding to the unfolded phase, and α, β, and γ are fitting parameters.

[0015] As a preferred embodiment of the present invention, the camera is an industrial camera.

[0016] As a preferred technical solution of the present invention, the industrial camera is a high-speed industrial camera.

[0017] As a preferred embodiment of the present invention, the projector is a DLP projector.

[0018] The beneficial effects of this invention are: the improved phase-height mapping method in 3D vision measurement based on hybrid device attitude requires only a small number of phase-height mapping formulas to achieve high-quality phase-height mapping, thus improving the practicality of phase-height mapping. Attached Figure Description

[0019] Figure 1 This is a diagram showing the effect of using the GRNN model to learn phase shift in this invention. Detailed Implementation

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1

[0022] The improved phase-height mapping method in 3D visual measurement based on hybrid device posture of the present invention is as follows: First, a projected fringe image is shown to the measurement target through a projector. Then, the unfolded phase obtained by analyzing the fringe image captured by the camera is analyzed. The unfolded phase is the difference between the unfolded phase with and without the measurement target. Then, the GRNN (Generalized Regressive Neural Network) model is used to reduce the phase information offset caused by the randomness of device posture at the same height. Finally, the unfolded phase with offset eliminated is substituted into the phase-height mapping formula to complete the mapping of phase information to height information.

[0023] The improved phase-height mapping method for 3D visual measurement based on hybrid device attitude of the present invention is implemented according to the following steps:

[0024] Step 1: Using formula (1), the coordinates (x, y) of the unfolded phase shift of the stripe image are taken as input, and the values ​​of the unfolded phase shift at the corresponding coordinates are taken as output, to train the GRNN model.

[0025] ψ offset =GRNN(x,y) (1)

[0026] In the formula: ψ offset To expand the phase offset at the corresponding coordinates;

[0027] Step 2, using formula (2), let ψ minus ψ offset The expanded phase, denoted as ψ′, is obtained after eliminating the effect of the expanded phase shift.

[0028] ψ′=ψ-ψ offset (2)

[0029] In the formula: ψ is the expanded phase;

[0030] Step 3: Substitute ψ′ as input into formula (3) to obtain a high-quality phase-height mapping result;

[0031]

[0032] In the formula: h(x,y) is the actual height value corresponding to the unfolded phase, and α, β, and γ are fitting parameters.

[0033] This invention, through analysis, reveals that in practical applications, the attitude of devices such as projectors and cameras in phase measurement profilometry systems can lead to phase shifts in the obtained unfolded phase information. This invention trains a GRNN model to obtain the offset value of the unfolded phase corresponding to each coordinate point. Then, it subtracts the offset value predicted by the GRNN model at the corresponding coordinates from the actual unfolded phase to obtain the unfolded phase that eliminates the phase shift effect. Finally, this phase information is substituted into the phase-height mapping formula to achieve high-quality phase-height mapping. This invention solves the problems of existing implicit phase-height mapping methods that do not consider the influence of device attitude and require a mapping formula for each pixel position, resulting in large memory consumption. This invention requires only a few phase-height mapping formulas to achieve high-quality phase-height mapping, improving the practicality of phase-height mapping.

[0034] Figure 1 The image shows the effect of the predicted unfolded phase offset value and the actual unfolded phase offset value of the trained GRNN model. As can be seen from Figure 1, the present invention can predict the phase offset value of each pixel well, which is very close to the actual phase offset value.

[0035] Example 2

[0036] The improved phase-height mapping method for 3D visual measurement based on hybrid device attitude of the present invention is implemented according to the following steps:

[0037] Step 1: Using formula (1), the coordinates (x, y) of the unfolded phase shift of the stripe image are taken as input, and the values ​​of the unfolded phase shift at the corresponding coordinates are taken as output, to train the GRNN model.

[0038] ψ offset =GRNN(x,y) (1)

[0039] In the formula: ψ offset To expand the phase offset at the corresponding coordinates;

[0040] Step 2, using formula (2), let ψ minus ψ offset The expanded phase, denoted as ψ′, is obtained after eliminating the effect of the expanded phase shift.

[0041] ψ′=ψ-ψ offset (2)

[0042] In the formula: ψ is the expanded phase;

[0043] Step 3: Substitute ψ′ as input into formula (3) to obtain a high-quality phase-height mapping result;

[0044]

[0045] In the formula: h(x,y) is the actual height value corresponding to the unfolded phase, and α, β, and γ are fitting parameters. In Example 2, the two sets of parameters that meet the requirements are: the first set: α is 0.002706, β is 0.4129, and γ is -0.5895; the second set: α is 0.005374, β is 0.4672, and γ is -0.36.

[0046] Example 3

[0047] Unlike Example 1, in the improved phase-height mapping method for 3D vision measurement based on hybrid device posture of the present invention, the industrial camera is a high-speed industrial camera and the projector is a DLP projector.

Claims

1. An improved phase-height mapping method for 3D visual measurement based on hybrid device attitude, characterized in that, An improved phase-height mapping method for 3D vision measurement based on hybrid device attitude is proposed. First, a fringe image is projected onto the surface of the target object using a projector. Then, the unfolded phase obtained from the fringe image captured by the camera is analyzed. The unfolded phase is the difference between the unfolded phase with and without the target object. Next, a GRNN model is used to reduce the phase information offset caused by the randomness of device attitude at the same height. Finally, the offset-free unfolded phase is substituted into the phase-height mapping formula to complete the mapping from phase information to height information. The specific steps are as follows: Step 1: Using formula (1), sequentially unfold the phase shift coordinates of the fringe image. As input, the expanded phase offset at the corresponding coordinates is used as output to train the GRNN model. (1) In the formula: To expand the phase offset at the corresponding coordinates; Step 2, using formula (2), let minus The unfolded phase, after eliminating the effect of unfolded phase shift, is obtained and denoted as... , (2) In the formula: To unfold the phase; Step 3, Substituting this into formula (3) yields a high-quality phase-height mapping result; (3) In the formula: To unfold the actual height value corresponding to the phase, α , β , γ These are the fitting parameters.

2. The improved phase-height mapping method for 3D visual measurement based on hybrid device attitude as described in claim 1, characterized in that, The camera in question is an industrial camera.

3. The improved phase-height mapping method for 3D visual measurement based on hybrid device attitude according to claim 2, characterized in that, The industrial camera is a high-speed industrial camera.

4. The improved phase-height mapping method for 3D visual measurement based on hybrid device attitude according to claim 3, characterized in that, The projector is a DLP projector.

Images