Satellite cabin plate assembly deviation detection method based on computer vision

Abstract

The application relates to the technical field of computer vision, and discloses a satellite cabin plate assembly deviation detection method based on computer vision, which comprises the following steps: collecting a polarization signal through a binocular camera and analyzing an initial discrete phase. Firstly, a polarization difference mapping diagram and a spatial partial derivative are calculated to obtain a surface local curvature. A phase offset compensation amount is calculated in combination with the curvature to obtain a compensation center phase. Then, a transient time phase variation rate is calculated, and a space-time phase change gradient is calculated. A reference projection tensor is calculated by using the gradient and a physical limit constraint coefficient, and a stable reference phase is extracted. Finally, an anisotropy correction factor is calculated based on this, a three-dimensional correction depth is reconstructed, and a final assembly deviation vector is obtained.

Description

Technical Field

[0001] This invention relates to the field of computer vision technology, specifically to a method for detecting satellite panel assembly deviations based on computer vision. Background Technology

[0002] In the field of spacecraft manufacturing, the assembly quality of satellite modules directly affects the structural stiffness and functional performance of the entire satellite. Currently, phase-shifting structured light or binocular vision technologies based on computer vision are commonly used for assembly deviation detection. These technologies reconstruct the three-dimensional shape by projecting and resolving surface grating fringes. However, in actual satellite module assembly scenarios, existing visual inspection technologies face several complex physical interferences. First, satellite module surfaces are typically covered with thermal control coatings such as polyimide films. These highly reflective materials are prone to severe specular highlight overflow under structured light projection. This leads to localized overexposure of the image, resulting in missing phase data at assembly gaps. Second, to suppress highlights, polarization filtering is often introduced into the inspection system. However, the module surface has complex geometric curvature variations. The uneven distribution of curvature causes non-uniform depolarization of polarized light. This causes a non-linear shift in the perceived phase center, introducing depth reconstruction errors. Furthermore, modules are subjected to continuous external stress during assembly and docking. Stress release causes transient micro-deformation jitter in the module structure. Existing static vision methods cannot effectively isolate dynamic displacement components in the time domain. This leads to the decoupling of the phase coordinates acquired in the previous moment from the current actual physical state. Furthermore, satellite panels often utilize honeycomb sandwich composite materials. This material exhibits significant anisotropy in deformation under stress, with differences in tensile deformation across different texture directions. Traditional isotropic reconstruction algorithms cannot adaptively correct for this. This leaves residual anisotropic spatial distortion in the 3D topography, resulting in distorted final detection results.

[0003] Therefore, a computer vision-based method for detecting satellite panel assembly deviations is needed to solve the problems existing in the prior art. Summary of the Invention

[0004] This invention provides a computer vision-based method for detecting satellite module assembly deviations, which helps to solve the problems mentioned in the background art.

[0005] This invention provides the following technical solution: a method for detecting assembly deviations of satellite panels based on computer vision, comprising: A detection system consisting of a phase-shifting structured light projector and binocular cameras equipped with zero-degree and ninety-degree polarizers is constructed. The phase-shifting structured light projector projects phase-shifting fringes onto the satellite module, and the binocular cameras are controlled to synchronously acquire optical signals. The horizontal polarization intensity, vertical polarization intensity, and initial discrete phase are extracted from the optical signals. The objective inherent structural parameters of the detection system, such as pixel physical size, binocular baseline length, and lens focal length, as well as the previous frame's compensated phase from the previous cycle's historical state data, are extracted, and a preset sampling time interval and theoretical assembly depth are input. A polarization difference mapping is calculated based on the horizontally polarized light intensity and the vertically polarized light intensity. The horizontal spatial partial derivative and the vertical spatial partial derivative are then calculated using the polarization difference mapping and the pixel physical size. The local curvature of the surface is calculated from the spatial partial derivatives in the horizontal and vertical directions. The phase offset compensation amount is calculated based on the local surface curvature, the pixel physical size, and the initial discrete phase. The compensation center phase is then calculated using the initial discrete phase and the phase offset compensation amount. The transient time phase variation rate is calculated by combining the compensation center phase, the compensation phase of the previous frame, and the sampling time interval; The spatiotemporal phase transition gradient is calculated using the transient time phase variation rate and the local surface curvature; The physical limit constraint coefficient is calculated based on the pixel physical size and the sampling time interval. The reference projection tensor is calculated using the spatiotemporal phase transition gradient and the physical limit constraint coefficient. The stable reference phase is calculated using the compensation center phase and the reference projection tensor. The anisotropic directional derivative is calculated based on the stable reference phase and the pixel physical size, and the anisotropic correction factor is calculated by combining the anisotropic directional derivative, the pixel physical size, and the stable reference phase. The three-dimensional correction depth is calculated based on the binocular baseline length, the lens focal length, the stable reference phase, the pixel physical size, and the anisotropic correction factor. The final assembly deviation vector is then calculated using the three-dimensional correction depth and the theoretical assembly depth.

[0006] Optionally, the step of calculating a polarization difference mapping map based on the horizontally polarized light intensity and the vertically polarized light intensity, and calculating the horizontal and vertical spatial partial derivatives using the polarization difference mapping map and the pixel physical size, includes: Obtain the horizontally polarized light intensity and the vertically polarized light intensity, and obtain the pixel physical size; Calculate the difference between the horizontally polarized light intensity and the vertically polarized light intensity; Calculate the sum of the horizontally polarized light intensity and the vertically polarized light intensity; Divide the difference by the sum to obtain the polarization difference mapping map; Calculate the difference between the polarization difference map shifted 1 pixel to the right and the polarization difference map shifted 1 pixel to the left; Divide the difference by twice the physical size of the pixel to obtain the spatial partial derivative in the horizontal direction; Calculate the difference between the polarization difference map offset downward by 1 pixel and the polarization difference map offset upward by 1 pixel; Divide the difference by twice the physical size of the pixel to obtain the spatial partial derivative in the vertical direction.

[0007] Optionally, calculating the local surface curvature from the spatial partial derivatives in the horizontal and vertical directions includes: Calculate the sum of the squares of the spatial partial derivatives in the horizontal direction and the squares of the spatial partial derivatives in the vertical direction; The local curvature of the surface is obtained by taking the square root of the sum of the squares.

[0008] Optionally, the step of calculating the phase offset compensation amount based on the local surface curvature, the pixel physical size, and the initial discrete phase, and calculating the compensation center phase using the initial discrete phase and the phase offset compensation amount, includes: Obtain the initial discrete phase; Calculate the product of the local curvature of the surface and the physical size of the pixel; Divide the product by the initial discrete phase to obtain the phase offset compensation amount; The sum of the phase offset compensation amount and the constant 1 is calculated and multiplied by the initial discrete phase to obtain the compensation center phase.

[0009] Optionally, the step of calculating the transient time phase change rate by combining the compensated center phase, the compensated phase of the previous frame, and the sampling time interval includes: Obtain the compensation phase of the previous frame and obtain the sampling time interval; Calculate the difference between the compensation center phase and the compensation phase of the previous frame; Dividing the difference by the sampling time interval yields the transient time phase variation rate.

[0010] Optionally, the step of calculating the spatiotemporal phase transition gradient using the transient time phase change rate and the local surface curvature includes: The spatiotemporal phase transition gradient is obtained by calculating the product of the transient time phase change rate and the local curvature of the surface.

[0011] Optionally, the step of calculating the physical limit constraint coefficient based on the pixel physical size and the sampling time interval, calculating the reference projection tensor using the spatiotemporal phase transition gradient and the physical limit constraint coefficient, and calculating the stable reference phase using the compensated center phase and the reference projection tensor includes: Calculate the reciprocal of the product of the pixel physical size and the sampling time interval to obtain the physical limit constraint coefficient; Calculate the sum of the spatiotemporal phase transition gradient and the physical limit constraint coefficient; Divide the spatiotemporal phase transition gradient by the sum to obtain the reference projection tensor; Calculate the difference between constant 1 and the reference projection tensor; Multiplying the compensation center phase by the difference yields a stable reference phase.

[0012] Optionally, the step of calculating the anisotropic directional derivative based on the stable reference phase and the pixel physical size, and calculating the anisotropic correction factor by combining the anisotropic directional derivative, the pixel physical size, and the stable reference phase, includes: Calculate the difference between the stable reference phase offset 1 pixel to the right and the stable reference phase at the center; Divide the difference by the pixel physical size to obtain the first ratio; Calculate the difference between the stable reference phase offset downward by 1 pixel and the stable reference phase at the center; Divide the difference by the pixel physical size to obtain the second ratio; Calculate the product of the first ratio and the second ratio to obtain the anisotropic directional derivative; Calculate the product of the anisotropic directional derivative and the square of the pixel physical size; Calculate the square of the stable reference phase; Dividing the product by the square of the stable reference phase yields the anisotropy correction factor.

[0013] Optionally, the step of calculating the three-dimensional correction depth based on the binocular baseline length, the lens focal length, the stable reference phase, the pixel physical size, and the anisotropy correction factor, and calculating the final assembly deviation vector using the three-dimensional correction depth and the theoretical assembly depth, includes: Obtain the binocular baseline length and the lens focal length, and obtain the theoretical assembly depth; Calculate the product of the binocular baseline length and the lens focal length to obtain the first product; Calculate the difference between constant 1 and the anisotropy correction factor; Calculate the product of the stable reference phase, the pixel physical size, and the difference to obtain the second product; Divide the first product by the second product to obtain the three-dimensional corrected depth. Calculate the difference between the three-dimensional correction depth and the theoretical assembly depth; Calculate the absolute value of the difference to obtain the final assembly deviation vector; The final assembly deviation vector is then output as the satellite panel assembly deviation detection result.

[0014] The present invention has the following beneficial effects: 1. By constructing a detection system including a phase-shifting structured light projector and a binocular camera to analyze the underlying optical signal, this scheme sequentially calculates the polarization difference map, local surface curvature, transient temporal phase change rate, spatiotemporal phase transition gradient, and anisotropic correction factor, thereby obtaining the three-dimensional correction depth and the final assembly deviation vector. In the specific environment of satellite module assembly, the specific material and spatial curvature of the module surface easily cause reflection interference and non-uniform depolarization distortion of the optical signal. Furthermore, the dynamic jitter caused by the release of assembly contact stress and the anisotropic stretching unique to the honeycomb interlayer material introduce nonlinear depth calculation errors. Considering this specific environment, this scheme uses the polarization difference map to extract the underlying contour to reduce reflection interference, adaptively compensates the initial discrete phase based on the local curvature, and calculates the reference projection tensor using the spatiotemporal phase transition gradient and physical limit constraint coefficient. This projects the dynamically deformed phase onto a stable reference frame, while simultaneously correcting the three-dimensional morphology using the anisotropic directional derivative. This approach reduces the adverse effects of reflection spillover, curvature distortion, transient jitter, and material anisotropy on the inspection, and can obtain a more accurate three-dimensional correction depth that reflects the true physical geometry, thereby improving the reliability of satellite panel assembly deviation inspection results.

[0015] 2. By acquiring the horizontal and vertical polarized light intensities, calculating their differences and sums, and performing division, a polarization difference mapping map is obtained. Then, combined with the camera's pixel physical size, the spatial partial derivatives in the horizontal and vertical directions are solved using the difference distribution of image pixels in orthogonal directions. This processing method effectively suppresses the common-mode high-light reflection component by utilizing the physical subtraction operation of orthogonal polarization states, reducing the specular overflow phenomenon caused by highly reflective thermal control coatings such as polyimide films on the satellite's surface. It converts the light intensity pixels originally affected by high-light interference into a dimensionless optical contrast distribution, extracting two-dimensional polarization difference mapping features that can reflect the contour of the physical surface, providing anti-interference photoelectric basic data for subsequent in-depth analysis of the spatial geometry of the satellite's surface.

[0016] 3. The local surface curvature is obtained by calculating the sum of the squares of the spatial partial derivatives in the horizontal direction and the square root of the sum of squares. This operation transforms the polarization difference mapping gradient tensor dispersed in two-dimensional orthogonal directions into a scalar curvature field. The Euclidean norm of the derivative objectively and directly characterizes the three-dimensional curvature of the satellite module material surface, establishing a quantitative mapping relationship between the underlying optical polarization partial derivatives and the macroscopic physical geometric curvature state. This eliminates the geometric perspective limitations caused by single-direction gradient evaluation, enabling the extracted curvature features to reflect the real undulations of the module surface and providing reliable spatial geometric curvature constraints for correcting subsequent non-uniform depolarization effects.

[0017] 4. By introducing an initial discrete phase, the product of the local curvature of the surface and the physical size of the pixel is calculated and divided by the initial discrete phase to obtain the phase offset compensation amount. Then, the phase offset compensation amount is summed with a set constant value and multiplied with the initial discrete phase to obtain the compensation center phase. This processing logic is based on the photophysical law that the greater the local curvature of the material surface, the more obvious the depolarization effect. The discrete phase calculated by the phase-shifting structured light is adaptively scaled using spatial geometric features to compensate for the depolarization distortion caused by the non-uniform curvature change of the satellite panel surface. This allows the sensed phase data center to return to the true physical contour position, reduces the nonlinear phase offset error caused by curvature degradation, and improves the objectivity of the single-frame phase morphology description.

[0018] 5. By acquiring the compensation phase of the previous frame and the set sampling time interval, the physical difference between the current compensation center phase and the compensation phase of the previous frame is calculated, and the difference is divided by the sampling time interval to obtain the transient time phase change rate. This operation uses the phase state difference between adjacent cycles and the system sampling frequency to extract the dynamic displacement features in the time domain, extending the static spatial phase to the time domain change dimension. It isolates the transient micro-deformation of the satellite panel caused by the continuous stress applied during the assembly process and the micro-jitter interference caused by stress release. It effectively separates the macroscopic static contour of the panel itself from the dynamic transient disturbance under the assembly cycle, and prepares a one-dimensional time change rate feature for the subsequent construction of joint frequency constraints and the removal of high-frequency displacement noise in the assembly.

[0019] 6. The spatiotemporal phase transition gradient is obtained by calculating the physical product of the transient time phase change rate and the local surface curvature. This operation orthogonally couples one-dimensional time-domain change rate information with two-dimensional spatial geometric curvature information to generate a joint gradient field feature that comprehensively covers time jitter and spatial deformation. This allows the spatiotemporal phase transition gradient to effectively distinguish between the global rigid translation of the cabin plate and the micro-deformation caused by local micro-stress. It achieves objective positioning of the stress concentration release area of ​​the assembly, overcomes the limitation that relying solely on time difference or spatial difference can easily confuse the interference source, and ensures that subsequent dynamic deformation filtering operations can be applied to the real micro-stress deformation concentration area in a targeted manner, thus enhancing the ability to identify interference under complex working conditions.

[0020] 7. The physical limit constraint coefficient is obtained by calculating the reciprocal of the product of the pixel physical size and the sampling time interval. The sum of the physical limit constraint coefficient and the spatiotemporal phase transition gradient is then calculated to obtain the reference projection tensor. Finally, the stable reference phase is extracted based on the product of the set constant, the difference between the reference projection tensor and the compensation center phase. This process uses the spatial resolution and temporal sampling frequency of the camera to construct a dimensionally homogeneous physical constraint threshold, avoiding the blind intervention of subjective weights. The constructed projection tensor smoothly projects the micro-deformation phase with dynamic jitter interference to a stable reference coordinate system, suppressing the high-frequency fluctuation noise caused by transient deformation during assembly, and restoring the output reference phase coordinates to a decoupled state bound to the real geometric state.

[0021] 8. The ratio of the stable reference phase in the orthogonal offset direction to the stable reference phase at the center is obtained by calculating the difference between the two directions and dividing by the pixel physical size. The ratio of the two is then multiplied to obtain the anisotropic directional derivative. The anisotropic correction factor is calculated by dividing the square of the pixel physical size and the square of the stable reference phase. This operation uses the orthogonal differential product of the phases of adjacent pixels to extract the cross derivative feature in physical space, quantifies the degree of anisotropic asymmetric stretching of the satellite cellular sandwich panel material under stress in different texture directions, and forms a distortion mapping field that can reflect the specificity of the material texture. This gives the system the ability to adaptively correct the morphology of anisotropic materials and provides a basis for directional distortion compensation for the final depth calibration.

[0022] 9. By acquiring the binocular baseline length, lens focal length, and the set theoretical assembly depth, the product of the binocular visual parameters is calculated as the first product. The second product is then calculated by combining the set constant value, stable reference phase, and pixel physical size. The two products are divided to obtain the three-dimensional corrected depth. Finally, the absolute value of the difference between the two products and the theoretical assembly depth is taken to obtain the final assembly deviation vector. This operation is based on the principle of triangulation and introduces anisotropic correction factors and spatial normalized size factors into the depth reconstruction process. While completing the homogeneous transformation of physical length dimensions, it corrects the residual anisotropic spatial distortion in different directions, reduces the local deformation distortion that is easily generated by direct reconstruction in traditional vision algorithms, and calculates and outputs the final deviation value that can characterize the current assembly quality. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the basic process of the present invention.

[0024] Figure 2 This is a schematic diagram showing the layout and parameter relationships of the detection system of the present invention.

[0025] Figure 3 This is a schematic diagram illustrating the relationship between polarization difference, spatial partial derivative, and phase compensation in this invention.

[0026] Figure 4 This is a schematic diagram illustrating the spatiotemporal phase transition constraint, anisotropy correction, and deviation output relationship of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Example 1, refer to Figure 1 A computer vision-based method for detecting assembly deviations in satellite cabin panels, comprising: A detection system consisting of a phase-shifting structured light projector and binocular cameras equipped with zero-degree and ninety-degree polarizers is constructed. The phase-shifting structured light projector projects phase-shifting fringes onto the satellite module, and the binocular cameras are controlled to synchronously acquire optical signals. The horizontal polarization intensity, vertical polarization intensity, and initial discrete phase are extracted from the optical signals. The objective inherent structural parameters of the detection system, such as pixel physical size, binocular baseline length, and lens focal length, as well as the previous frame's compensated phase from the previous cycle's historical state data, are extracted, and a preset sampling time interval and theoretical assembly depth are input. Reference Figure 2The detection system consists of a phase-shifting structured light projector and two sets of polarized binocular cameras arranged around the satellite module, and collects optical analysis results, equipment structural parameters and historical design parameters accordingly. A polarization difference mapping is calculated based on the horizontally polarized light intensity and the vertically polarized light intensity. The horizontal spatial partial derivative and the vertical spatial partial derivative are then calculated using the polarization difference mapping and the pixel physical size. The local curvature of the surface is calculated from the spatial partial derivatives in the horizontal and vertical directions. The phase offset compensation amount is calculated based on the local surface curvature, the pixel physical size, and the initial discrete phase. The compensation center phase is then calculated using the initial discrete phase and the phase offset compensation amount. The transient time phase variation rate is calculated by combining the compensation center phase, the compensation phase of the previous frame, and the sampling time interval; The spatiotemporal phase transition gradient is calculated using the transient time phase variation rate and the local surface curvature; The physical limit constraint coefficient is calculated based on the pixel physical size and the sampling time interval. The reference projection tensor is calculated using the spatiotemporal phase transition gradient and the physical limit constraint coefficient. The stable reference phase is calculated using the compensation center phase and the reference projection tensor. The anisotropic directional derivative is calculated based on the stable reference phase and the pixel physical size, and the anisotropic correction factor is calculated by combining the anisotropic directional derivative, the pixel physical size, and the stable reference phase. The three-dimensional correction depth is calculated based on the binocular baseline length, the lens focal length, the stable reference phase, the pixel physical size, and the anisotropic correction factor. The final assembly deviation vector is then calculated using the three-dimensional correction depth and the theoretical assembly depth.

[0029] The step of calculating a polarization difference mapping map based on the horizontally polarized light intensity and the vertically polarized light intensity, and then using the polarization difference mapping map and the pixel physical size to calculate the horizontal and vertical spatial partial derivatives, includes: The purpose and principle of this step is to solve the problem of specular highlight overflow caused by highly reflective coatings; by performing differential operations on orthogonal polarization images, common-mode specular reflection components are eliminated, and the polarization difference mapping map of the physical contour is extracted. Obtain the intensity of horizontally polarized light and the intensity of vertically polarized light, and at the same time obtain the physical size of the camera's pixels; First, to extract the underlying contour features without specular interference, a polarization difference mapping map is calculated through algebraic operations on the horizontal and vertical polarization light intensities. The calculation steps are as follows: Calculate the difference between the horizontally polarized light intensity and the vertically polarized light intensity; calculate the sum of the horizontally polarized light intensity and the vertically polarized light intensity; divide the difference by the sum to obtain the polarization difference mapping map; Next, to quantify the geometric features of the cabin surface, the physical space partial derivative of the polarization difference mapping map in the horizontal direction is calculated. The calculation steps are as follows: Calculate the difference between the polarization difference map offset by 1 pixel to the right and the polarization difference map offset by 1 pixel to the left; divide the difference by twice the physical size of the pixel to obtain the spatial partial derivative in the horizontal direction. Next, the physical space partial derivatives of the polarization difference mapping diagram in the vertical direction are calculated. The calculation steps are as follows: Calculate the difference between the polarization difference map offset downward by 1 pixel and the polarization difference map offset upward by 1 pixel; divide the difference by twice the physical size of the pixel to obtain the spatial partial derivative in the vertical direction.

[0030] By acquiring the horizontal and vertical polarized light intensities, calculating their differences and sums, and then dividing them, a polarization difference mapping map is obtained. This map is then combined with the camera's pixel physical dimensions, and the spatial partial derivatives in the horizontal and vertical directions are calculated using the difference distribution of image pixels in orthogonal directions. This processing method effectively suppresses common-mode high-reflection components through physical subtraction of orthogonal polarization states, reducing the specular overflow phenomenon caused by highly reflective thermal control coatings such as polyimide films on the satellite module surface. It converts the light intensity pixels originally affected by high-reflection interference into a dimensionless optical contrast distribution, extracting two-dimensional polarization difference mapping features that reflect the contours of the physical surface. This provides interference-resistant photoelectric fundamental data for subsequent in-depth analysis of the module's spatial geometry.

[0031] The calculation of the local surface curvature from the spatial partial derivatives in the horizontal and vertical directions includes: The purpose and principle of this step is to transform the output two-dimensional gradient tensor into a scalar curvature field; and to objectively characterize the degree of three-dimensional curling of the material surface by using the square root of the sum of the squares of the spatial partial derivatives, thus providing a spatial geometric basis for subsequent correction of non-uniform depolarization effects. To quantify the severity of three-dimensional curling on the surface of the cabin plate, the local curvature of the surface is synthesized by combining horizontal and vertical partial derivatives. The calculation steps are as follows: Calculate the sum of the squares of the spatial partial derivatives in the horizontal direction and the squares of the spatial partial derivatives in the vertical direction; take the square root of the sum of squares to obtain the local curvature of the surface.

[0032] The local surface curvature is obtained by calculating the sum of the squares of the spatial partial derivatives in the horizontal direction and the square root of the sum of squares. This operation transforms the polarization difference mapping gradient tensor dispersed in two-dimensional orthogonal directions into a scalar curvature field. The Euclidean norm of the derivative objectively and directly characterizes the three-dimensional curvature of the satellite panel material surface, establishing a quantitative mapping relationship between the underlying optical polarization partial derivatives and the macroscopic physical geometric curvature state. This eliminates the geometric perspective limitations caused by single-direction gradient evaluation, enabling the extracted curvature features to reflect the real undulations of the panel surface and providing reliable spatial geometric curvature constraints for correcting subsequent non-uniform depolarization effects.

[0033] The step of calculating the phase offset compensation amount based on the local surface curvature, the pixel physical size, and the initial discrete phase, and calculating the compensation center phase using the initial discrete phase and the phase offset compensation amount, includes: Reference Figure 3 The horizontal polarization intensity, vertical polarization intensity, and pixel physical size are used to form a polarization difference map, and further correlated with spatial partial derivatives, local surface curvature, and phase compensation parameters. The purpose and principle of this step is to correct the phase center shift caused by the depolarization effect due to curvature; based on the rule that the greater the local curvature, the more obvious the phase shift, the discrete phase is adaptively proportionally compensated by the local curvature of the surface. Obtain the initial discrete phase obtained by objective photoelectric calculation of the system; First, to determine the degree of phase degradation caused by surface curvature, the phase offset compensation is calculated. The calculation steps are as follows: Calculate the product of the local surface curvature and the pixel physical size; divide the product by the initial discrete phase to obtain the phase offset compensation amount; Subsequently, this compensation amount is superimposed on the original data, and the compensation center phase of the current frame is calculated. The calculation steps are as follows: Calculate the sum of the phase offset compensation amount and the constant 1; multiply the sum by the initial discrete phase to obtain the compensation center phase.

[0034] By introducing an initial discrete phase, the product of the local curvature of the surface and the physical size of the pixel is calculated and divided by the initial discrete phase to obtain the phase offset compensation amount. Then, the phase offset compensation amount is summed with a set constant value and multiplied with the initial discrete phase to obtain the compensation center phase. This processing logic is based on the photophysical law that the greater the local curvature of the material surface, the more obvious the depolarization effect. The discrete phase calculated by the phase-shifting structured light is adaptively scaled using spatial geometric features to compensate for the depolarization distortion caused by the non-uniform curvature change of the satellite panel surface. This allows the sensed phase data center to return to the true physical contour position, reduces the nonlinear phase offset error caused by curvature degradation, and improves the objectivity of single-frame phase morphology description.

[0035] The calculation of the transient time phase change rate by combining the compensated center phase, the compensated phase of the previous frame, and the sampling time interval includes: The purpose and principle of this step is to initially quantify the transient micro-deformation of the cabin plate caused by the stress release during assembly; by extracting the rate of change of the difference between the current compensation phase and the phase of the previous sampling period with respect to time, the dynamic displacement components in the time domain are effectively isolated. Obtain the compensation phase of the previous frame. When the current frame is the first acquisition frame, use the compensation center phase of the current frame as the initial value of the compensation phase of the previous frame. Get the set sampling time interval; To capture the dynamic characteristics of transient displacement during the assembly process, the transient time phase change rate is calculated. The calculation steps are as follows: Calculate the difference between the compensated center phase and the compensated phase of the previous frame; divide the difference by the sampling time interval to obtain the transient time phase variation rate.

[0036] By acquiring the compensation phase of the previous frame and the set sampling time interval, the physical difference between the current compensation center phase and the compensation phase of the previous frame is calculated, and the difference is divided by the sampling time interval to obtain the transient time phase variation rate. This operation uses the phase state difference between adjacent cycles and the system sampling frequency to extract the dynamic displacement features in the time domain, extending the static spatial phase to the time domain change dimension. It isolates the transient micro-deformation of the satellite panel caused by the continuous stress applied during the assembly process and the micro-jitter interference caused by stress release. It effectively separates the macroscopic static contour of the panel itself from the dynamic transient disturbance under the assembly cycle, and prepares a one-dimensional time variation rate feature for the subsequent construction of joint frequency constraints and the removal of high-frequency displacement noise in the assembly.

[0037] The calculation of the spatiotemporal phase transition gradient using the transient time phase change rate and the local surface curvature includes: Reference Figure 4The compensation center phase is processed by time variation and physical limit constraints to form a stable reference phase, which is then connected with anisotropic correction, three-dimensional correction depth and deviation output. The function and principle of this step is to orthogonally couple the one-dimensional time variation rate with the local curvature in space; the resulting gradient field can locate the concentrated area of ​​local deformation caused by micro-stress and eliminate the interference of global rigid translation. To locate the local micro-stress deformation region, construct the overall spatiotemporal phase transition characteristics, and calculate the spatiotemporal phase transition gradient, the calculation steps are as follows: The spatiotemporal phase transition gradient is obtained by calculating the product of the transient time phase change rate and the local curvature of the surface.

[0038] The spatiotemporal phase transition gradient is obtained by calculating the physical product of the transient temporal phase change rate and the local curvature of the surface. This operation orthogonally couples one-dimensional temporal change rate information with two-dimensional spatial geometric curvature information to generate a joint gradient field feature that comprehensively covers temporal jitter and spatial deformation. This allows the spatiotemporal phase transition gradient to effectively distinguish between the global rigid translation of the panel and the micro-deformation caused by local micro-stress, thus achieving objective positioning of the stress concentration release area in the assembly. This overcomes the limitation that relying solely on temporal or spatial difference can easily confuse interference sources, ensuring that subsequent dynamic deformation filtering operations can be applied specifically to the real micro-stress deformation concentration area, and enhancing the ability to identify interference under complex working conditions.

[0039] The step of calculating the physical limit constraint coefficient based on the pixel physical size and the sampling time interval, calculating the reference projection tensor using the spatiotemporal phase transition gradient and the physical limit constraint coefficient, and calculating the stable reference phase using the compensated center phase and the reference projection tensor includes: The purpose and principle of this step is to project the dynamic phase with slight deformation onto a stable reference coordinate system; by constructing physical limit constraint coefficients that are homogeneous with the spatiotemporal phase transition gradient, the dimensionless projection tensor is obtained to suppress dynamic deformation noise. First, based on the camera's physical sampling limit, the physical limit constraint coefficient is calculated. The calculation steps are as follows: Calculate the reciprocal of the product of the pixel physical size and the sampling time interval to obtain the physical limit constraint coefficient; Next, based on the phase transition gradient and limit constraints, the reference projection tensor is calculated. The calculation steps are as follows: Calculate the sum of the spatiotemporal phase transition gradient and the physical limit constraint coefficient; divide the spatiotemporal phase transition gradient by the sum to obtain the reference projection tensor; Then, the projection tensor is used to suppress transient deformation disturbances and obtain a stable reference phase. The calculation steps are as follows: Calculate the difference between constant 1 and the reference projection tensor; multiply the compensation center phase by the difference to obtain the stable reference phase.

[0040] The physical limit constraint coefficient is obtained by calculating the reciprocal of the product of the pixel physical size and the sampling time interval. The sum of the physical limit constraint coefficient and the spatiotemporal phase transition gradient is then calculated to obtain the reference projection tensor. Finally, the stable reference phase is extracted based on the product of the difference between the set constant and the reference projection tensor and the compensation center phase. This process uses the spatial resolution of the camera and the temporal sampling frequency to construct a dimensionally homogeneous physical constraint threshold, avoiding the blind intervention of subjective weights. The constructed projection tensor smoothly projects the micro-deformation phase with dynamic jitter interference to a stable reference coordinate system, suppressing the high-frequency fluctuation noise caused by transient deformation during assembly. This restores the output reference phase coordinates to a decoupled state bound to the real geometric state.

[0041] The step of calculating the anisotropic directional derivative based on the stable reference phase and the pixel physical size, and then calculating the anisotropic correction factor by combining the anisotropic directional derivative, the pixel physical size, and the stable reference phase, includes: The purpose and principle of this step is to extract the distortion mapping field that characterizes the anisotropic features of the honeycomb sandwich material; by calculating the cross derivative of the stable reference phase in physical space, the subsequent depth calculation has the ability to adaptively correct for stretching of the material in different texture directions. First, the anisotropic directional derivative is calculated by the orthogonal difference product of the phases of adjacent pixels. The calculation steps are as follows: Calculate the difference between the stable reference phase offset 1 pixel to the right and the stable reference phase at the center; divide the difference by the pixel physical size to obtain a first ratio; calculate the difference between the stable reference phase offset 1 pixel downward and the stable reference phase at the center; divide the difference by the pixel physical size to obtain a second ratio; calculate the product of the first ratio and the second ratio to obtain the anisotropic directional derivative. Next, based on the ratio of the directional derivative to the stable phase, the anisotropy correction factor is calculated. The calculation steps are as follows: Calculate the product of the anisotropic directional derivative and the square of the pixel physical size; calculate the square of the stable reference phase; divide the product by the square of the stable reference phase to obtain the anisotropic correction factor.

[0042] By calculating the difference between the stable reference phase in the orthogonal offset direction and the stable reference phase at the center, and dividing it by the pixel physical size, the ratio of the two directions is obtained. Multiplying the two yields the anisotropic directional derivative. Then, the anisotropic correction factor is calculated by dividing the square of the pixel physical size and the square of the stable reference phase. This operation extracts the cross-derivative feature in physical space using the orthogonal differential product of adjacent pixel phases. It quantifies the degree of anisotropic asymmetric stretching of the satellite cellular sandwich panel material under stress in different texture directions, forming a distortion mapping field that reflects the material texture specificity. This endows the system with adaptive morphology correction capability for anisotropic materials and provides a basis for directional distortion compensation for the final depth calibration.

[0043] The step of calculating the three-dimensional correction depth based on the binocular baseline length, the lens focal length, the stable reference phase, the pixel physical size, and the anisotropy correction factor, and then calculating the final assembly deviation vector using the three-dimensional correction depth and the theoretical assembly depth, includes: The purpose of this step is to combine the anisotropic correction factor and introduce the spatial normalized size factor to calculate the three-dimensional depth in the physical coordinate system and obtain the target deviation data that reflects the assembly quality. Obtain the binocular baseline length and lens focal length; Obtain the set theoretical assembly depth; First, by combining binocular vision parameters, spatial pixel normalization factor, and anisotropic correction factor, the 3D corrected depth is reconstructed. The calculation steps are as follows: Calculate the product of the binocular baseline length and the lens focal length to obtain a first product; calculate the difference between the constant 1 and the anisotropy correction factor; calculate the product of the stable reference phase, the pixel physical size and the difference to obtain a second product; divide the first product by the second product to obtain the three-dimensional correction depth; Finally, by reconstructing the absolute difference between the depth and the theoretical design data, the final assembly deviation vector is calculated. The calculation steps are as follows: Calculate the difference between the three-dimensional correction depth and the theoretical assembly depth; calculate the absolute value of the difference to obtain the final assembly deviation vector; The final assembly deviation vector is then output as the satellite panel assembly deviation detection result.

[0044] By acquiring the binocular baseline length, lens focal length, and a set theoretical assembly depth, the product of the binocular visual parameters is calculated as the first product. The second product is then calculated by combining the set constant value, stable reference phase, and pixel physical size. The two products are divided to obtain the three-dimensional corrected depth. Finally, the absolute value of the difference between the corrected depth and the theoretical assembly depth is taken to obtain the final assembly deviation vector. This operation is based on the principle of triangulation and introduces anisotropic correction factors and spatial normalized size factors into the depth reconstruction process. While completing the homogeneous transformation of physical length dimensions, it corrects the residual anisotropic spatial distortion in different directions, reduces the local deformation distortion that is easily generated by direct reconstruction in traditional vision algorithms, and calculates and outputs the final deviation value that can characterize the current assembly quality.

[0045] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0046] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for detecting assembly deviations of satellite cabin panels based on computer vision, characterized in that, include: A detection system consisting of a phase-shifting structured light projector and binocular cameras equipped with zero-degree and ninety-degree polarizers is constructed. The phase-shifting structured light projector projects phase-shifting fringes onto the satellite module, and the binocular cameras are controlled to synchronously acquire optical signals. The horizontal polarization intensity, vertical polarization intensity, and initial discrete phase are extracted from the optical signals. The objective inherent structural parameters of the detection system, such as pixel physical size, binocular baseline length, and lens focal length, as well as the previous frame's compensated phase from the previous cycle's historical state data, are extracted, and a preset sampling time interval and theoretical assembly depth are input. A polarization difference mapping is calculated based on the horizontally polarized light intensity and the vertically polarized light intensity. The horizontal spatial partial derivative and the vertical spatial partial derivative are then calculated using the polarization difference mapping and the pixel physical size. The local curvature of the surface is calculated from the spatial partial derivatives in the horizontal and vertical directions. The phase offset compensation amount is calculated based on the local surface curvature, the pixel physical size, and the initial discrete phase. The compensation center phase is then calculated using the initial discrete phase and the phase offset compensation amount. The transient time phase variation rate is calculated by combining the compensation center phase, the compensation phase of the previous frame, and the sampling time interval; The spatiotemporal phase transition gradient is calculated using the transient time phase variation rate and the local surface curvature; The physical limit constraint coefficient is calculated based on the pixel physical size and the sampling time interval. The reference projection tensor is calculated using the spatiotemporal phase transition gradient and the physical limit constraint coefficient. The stable reference phase is calculated using the compensation center phase and the reference projection tensor. The anisotropic directional derivative is calculated based on the stable reference phase and the pixel physical size, and the anisotropic correction factor is calculated by combining the anisotropic directional derivative, the pixel physical size, and the stable reference phase. The three-dimensional correction depth is calculated based on the binocular baseline length, the lens focal length, the stable reference phase, the pixel physical size, and the anisotropic correction factor. The final assembly deviation vector is then calculated using the three-dimensional correction depth and the theoretical assembly depth.

2. The satellite module assembly deviation detection method based on computer vision according to claim 1, characterized in that, The step of calculating a polarization difference mapping map based on the horizontally polarized light intensity and the vertically polarized light intensity, and then using the polarization difference mapping map and the pixel physical size to calculate the horizontal and vertical spatial partial derivatives, includes: Obtain the horizontally polarized light intensity and the vertically polarized light intensity, and obtain the pixel physical size; Calculate the difference between the horizontally polarized light intensity and the vertically polarized light intensity; Calculate the sum of the horizontally polarized light intensity and the vertically polarized light intensity; Divide the difference by the sum to obtain the polarization difference mapping map; Calculate the difference between the polarization difference map shifted 1 pixel to the right and the polarization difference map shifted 1 pixel to the left; Divide the difference by twice the physical size of the pixel to obtain the spatial partial derivative in the horizontal direction; Calculate the difference between the polarization difference map offset downward by 1 pixel and the polarization difference map offset upward by 1 pixel; Divide the difference by twice the physical size of the pixel to obtain the spatial partial derivative in the vertical direction.

3. The satellite module assembly deviation detection method based on computer vision according to claim 2, characterized in that, The calculation of the local surface curvature from the spatial partial derivatives in the horizontal and vertical directions includes: Calculate the sum of the squares of the spatial partial derivatives in the horizontal direction and the squares of the spatial partial derivatives in the vertical direction; The local curvature of the surface is obtained by taking the square root of the sum of the squares.

4. The satellite module assembly deviation detection method based on computer vision according to claim 3, characterized in that, The step of calculating the phase offset compensation amount based on the local surface curvature, the pixel physical size, and the initial discrete phase, and calculating the compensation center phase using the initial discrete phase and the phase offset compensation amount, includes: Obtain the initial discrete phase; Calculate the product of the local curvature of the surface and the physical size of the pixel; Divide the product by the initial discrete phase to obtain the phase offset compensation amount; The sum of the phase offset compensation amount and the constant 1 is calculated and multiplied by the initial discrete phase to obtain the compensation center phase.

5. The satellite module assembly deviation detection method based on computer vision according to claim 4, characterized in that, The calculation of the transient time phase change rate by combining the compensated center phase, the compensated phase of the previous frame, and the sampling time interval includes: Obtain the compensation phase of the previous frame and obtain the sampling time interval; Calculate the difference between the compensation center phase and the compensation phase of the previous frame; Dividing the difference by the sampling time interval yields the transient time phase variation rate.

6. The satellite module assembly deviation detection method based on computer vision according to claim 5, characterized in that, The calculation of the spatiotemporal phase transition gradient using the transient time phase change rate and the local surface curvature includes: The spatiotemporal phase transition gradient is obtained by calculating the product of the transient time phase change rate and the local curvature of the surface.

7. The satellite module assembly deviation detection method based on computer vision according to claim 6, characterized in that, The step of calculating the physical limit constraint coefficient based on the pixel physical size and the sampling time interval, calculating the reference projection tensor using the spatiotemporal phase transition gradient and the physical limit constraint coefficient, and calculating the stable reference phase using the compensated center phase and the reference projection tensor includes: Calculate the reciprocal of the product of the pixel physical size and the sampling time interval to obtain the physical limit constraint coefficient; Calculate the sum of the spatiotemporal phase transition gradient and the physical limit constraint coefficient; Divide the spatiotemporal phase transition gradient by the sum to obtain the reference projection tensor; Calculate the difference between constant 1 and the reference projection tensor; Multiplying the compensation center phase by the difference yields a stable reference phase.

8. The satellite module assembly deviation detection method based on computer vision according to claim 7, characterized in that, The step of calculating the anisotropic directional derivative based on the stable reference phase and the pixel physical size, and then calculating the anisotropic correction factor by combining the anisotropic directional derivative, the pixel physical size, and the stable reference phase, includes: Calculate the difference between the stable reference phase offset 1 pixel to the right and the stable reference phase at the center; Divide the difference by the pixel physical size to obtain the first ratio; Calculate the difference between the stable reference phase offset downward by 1 pixel and the stable reference phase at the center; Divide the difference by the pixel physical size to obtain the second ratio; Calculate the product of the first ratio and the second ratio to obtain the anisotropic directional derivative; Calculate the product of the anisotropic directional derivative and the square of the pixel physical size; Calculate the square of the stable reference phase; Dividing the product by the square of the stable reference phase yields the anisotropy correction factor.

9. A method for detecting satellite module assembly deviation based on computer vision according to claim 8, characterized in that, The step of calculating the three-dimensional correction depth based on the binocular baseline length, the lens focal length, the stable reference phase, the pixel physical size, and the anisotropy correction factor, and then calculating the final assembly deviation vector using the three-dimensional correction depth and the theoretical assembly depth, includes: Obtain the binocular baseline length and the lens focal length, and obtain the theoretical assembly depth; Calculate the product of the binocular baseline length and the lens focal length to obtain the first product; Calculate the difference between constant 1 and the anisotropy correction factor; Calculate the product of the stable reference phase, the pixel physical size, and the difference to obtain the second product; Divide the first product by the second product to obtain the three-dimensional corrected depth. Calculate the difference between the three-dimensional correction depth and the theoretical assembly depth; Calculate the absolute value of the difference to obtain the final assembly deviation vector; The final assembly deviation vector is then output as the satellite panel assembly deviation detection result.

Images