CN-121994169-A - Laser microscopic three-dimensional roughness measuring method

Abstract

The invention relates to the technical field of laser measurement and material characterization, in particular to a laser microscopic three-dimensional roughness measurement method, which comprises the steps of coaxially spreading and irradiating a sample by multi-band ultrafast light pulse and picosecond ultrasonic pulse, and generating a density map by compressing and collecting interference and phase shift data by an event sensor; the neural wavefront prediction outputs a phase compensation matrix and an uncertainty tensor, a photoacoustic forward model is built by combining a water-containing refractive index field, a high voxel and an elastic voxel are rebuilt by an alternating direction multiplier algorithm under the constraint of contact coupling, fractal roughness, power spectral density and topological characteristics are extracted by cross-map convolution, a performance index is given by a multitask model, a performance gradient heat map is generated, the uncertainty and heat map feedback self-adaptively update sampling density and regularization parameters, three-dimensional roughness is output in a closed loop mode, and online detection is realized.

Inventors

• WANG SHUANBAO
• HU MIAOMIAO
• ZHAI XIAOJING
• ZENG CHENG
• FAN XIAOYI
• TANG JIAJIA
• LI MENGNAN
• WANG XINGGUANG
• LIU XUERONG
• HE ZHENKUN
• ZHANG QINGYU
• YAN XIAOMING
• DONG XIAO
• LI HUIXUN
• ZHAO YI
• An Jiexin

Assignees

• 石家庄市公路桥梁建设集团有限公司
• 河北交通职业技术学院
• 重庆交通大学

Dates

Publication Date

20260508

Application Date

20251023

Claims (10)

1. 1. The laser microscopic three-dimensional roughness measuring method is characterized by comprising the following steps of: the method comprises the steps of transmitting multi-band ultrafast optical pulses and picosecond ultrasonic pulses to coaxially combine beams, loading phases by a spatial light modulator to scan samples along a spiral after pseudo-random spread spectrum, collecting speckle event streams by an event sensor, compressing signals to obtain interference data, phase shift data and a sampling density map, and generating a risk field according to water content distribution; The method comprises the steps of inputting a speckle event stream into a nerve wavefront predictor, outputting a phase compensation matrix to refresh a spatial light modulator and generating an uncertainty tensor, constructing a refractive index field according to moisture content distribution, establishing an optical and acoustic forward operator, and solving an augmentation Lagrangian function under contact coupling constraint through an iterative algorithm to obtain a three-dimensional altitude voxel matrix and a local elastic voxel matrix; The method comprises the steps of constructing a morphology graph and an elastic graph by a three-dimensional high voxel matrix and a local elastic voxel matrix, extracting roughness, power spectrum and topological characteristics by cross-graph convolution, inputting the characteristics into a multi-task model to obtain roughness related performance indexes, generating a performance gradient heat graph and an uncertainty tensor, adjusting a sampling density graph according to the uncertainty tensor, updating regularization parameters of an augmented Lagrange function according to the performance gradient heat graph, and outputting a cement surface three-dimensional roughness result.
2. 2. The method of claim 1, wherein the multi-band ultrafast optical pulses and the picosecond ultrasonic pulses are spread spectrum encoded using the same pseudo-random sequence and propagate along the same optical axis.
3. 3. The method of claim 1 wherein the phase map loaded by the spatial light modulator is a bessel laguerre complex phase, the phase map having both radial bessel components and angular laguerre components.
4. 4. The method of claim 1, wherein the event sensor employs a pulsed intensity jump detection scheme, the intensity jump threshold being set by a current background noise mean value prior to each scan.
5. 5. The method of claim 1, wherein the leaky integrate transmit encoding layer of the neural wavefront predictor converts the event stream into discrete pulse tensors, and the axial attention convolution layer performs convolution and attention weight update in row and column directions, respectively.
6. 6. The method of claim 1, wherein the optical forward operator is established using a phase retardation kernel function of the refractive index field and the acoustic forward operator is established using a surface acoustic wave propagation kernel function.
7. 7. The method of claim 1, wherein the iterative algorithm is an alternating direction multiplier algorithm and the augmented lagrangian penalty factor is recalculated based on the main variable residuals at each iteration step.
8. 8. The method of claim 1, wherein the contact coupling constraint uses a hertz contact relationship to set the gradient of the local elastin matrix to the three-dimensional altitude voxel matrix to a fixed three-half index relationship.
9. 9. The method of claim 1, wherein an encoder decoder architecture is employed across graph convolution, wherein the encoding stage sets two-level graph convolution downsampling and retains a graph topology index, and wherein the decoding stage sets two-level graph convolution upsampling and concatenates corresponding encoding stage features.
10. 10. The method of claim 1, wherein the roughness-related performance index consists of a fractal roughness feature, a power spectral density feature, and a persistent coherent topology feature, wherein the performance gradient thermal map updates a sampling density map based on node gradient magnitudes, and wherein the uncertainty tensor is used to adjust the light pulse scan step size according to a risk field algorithm.

Description

Laser microscopic three-dimensional roughness measuring method Technical Field The invention relates to the technical field of laser measurement and material characterization, in particular to a laser microscopic three-dimensional roughness measurement method. Background The surface roughness of the cement product directly influences the adhesion, frost peeling resistance and light scattering performance of the coating, and is a core index for evaluating the durability of the concrete member. The traditional laboratory roughness measurement mostly adopts a confocal scanning microscope or a white light interferometer, wherein the confocal method needs a pinhole to intercept off-axis scattering, deep hole information is lost, and the white light interferometer has phase wrapping on a high-slope wall surface. In order to realize online detection of the production line, a single-wavelength laser triangulation scheme is provided, but the refractive index of the scheme is changed severely under the gradient environment of the water content, so that the axial error is obviously increased. The prior research attempts to superimpose ultrasonic echo into an optical altitude map and improve deep hole imaging by simple plane stitching, but photoacoustic asynchronous triggering generates time base deviation, so that coupling inversion is difficult to converge. In addition, the frame acquisition mode of the area-array camera has huge data redundancy under the pulse frequency of 80MHz, and is difficult to meet the real-time requirement. Disclosure of Invention Aiming at a plurality of problems existing in the prior art, the invention provides a laser microscopic three-dimensional roughness measurement method, which utilizes coaxial spread spectrum light-sound pulse to synchronously collect speckle event flow, simultaneously reconstructs three-dimensional height and elastic voxels through a nerve wave front prediction and photoacoustic coupling iteration solver, extracts roughness multi-scale characteristics through cross-map convolution, dynamically adjusts scanning density and regular parameters through performance gradient and uncertainty dual feedback, and finally outputs nano-scale cement roughness. A laser microscopic three-dimensional roughness measuring method comprises the following steps: the method comprises the steps of transmitting multi-band ultrafast optical pulses and picosecond ultrasonic pulses to coaxially combine beams, loading phases by a spatial light modulator to scan samples along a spiral after pseudo-random spread spectrum, collecting speckle event streams by an event sensor, compressing signals to obtain interference data, phase shift data and a sampling density map, and generating a risk field according to water content distribution; The method comprises the steps of inputting a speckle event stream into a nerve wavefront predictor, outputting a phase compensation matrix to refresh a spatial light modulator and generating an uncertainty tensor, constructing a refractive index field according to moisture content distribution, establishing an optical and acoustic forward operator, and solving an augmentation Lagrangian function under contact coupling constraint through an iterative algorithm to obtain a three-dimensional altitude voxel matrix and a local elastic voxel matrix; The method comprises the steps of constructing a morphology graph and an elastic graph by a three-dimensional high voxel matrix and a local elastic voxel matrix, extracting roughness, power spectrum and topological characteristics by cross-graph convolution, inputting the characteristics into a multi-task model to obtain roughness related performance indexes, generating a performance gradient heat graph and an uncertainty tensor, adjusting a sampling density graph according to the uncertainty tensor, updating regularization parameters of an augmented Lagrange function according to the performance gradient heat graph, and outputting a cement surface three-dimensional roughness result. Preferably, the multiband ultrafast optical pulses and picosecond ultrasonic pulses are spread spectrum coded using the same pseudo-random sequence and propagated along the same optical axis. Preferably, the phase map loaded by the spatial light modulator is a Bessel lager composite phase, and the phase map has both radial Bessel components and angular Laguerre components. Preferably, the event sensor adopts a pulse brightness jump detection mode, and the brightness jump threshold value is set by the current background noise mean value before each scanning. Preferably, the leaky integrate transmit encoding layer of the neural wavefront predictor converts the event stream into discrete pulse tensors, and the axial attention convolution layer performs convolution and attention weight update in row and column directions, respectively. Preferably, the optical forward operator is built using a phase retardation kernel function of the refractive index