CN-122017794-A - Visibility inversion method and system for multi-parameter fusion of light quantum radar

CN122017794ACN 122017794 ACN122017794 ACN 122017794ACN-122017794-A

Abstract

The invention relates to the technical field of photoelectric detection and remote sensing, and discloses a visibility inversion method and a visibility inversion system for multi-parameter fusion of a light quantum radar. According to the method, arrival time, spatial position, polarization state and wave front phase information of echo photons are synchronously acquired through single photon detection, an intensity scintillation index, a beam drift characteristic, a polarization evolution parameter and a Zernike low-order phase distortion coefficient are respectively extracted, a multi-parameter fusion characteristic vector is constructed, a pre-trained encoder-decoder neural network model is input, and high-precision visibility inversion is achieved. The system comprises a single photon detection module, a multidimensional feature extraction module and a visibility inversion module. The method separates turbulence interference and real visibility change, has small inversion error under strong turbulence, and is superior to the traditional method.

Inventors

WANG ZHUANG
HUO YANFENG
LIU CHENGXIAO
TAO YIN
ZHU YONG
YANG GUANYING
CHEN JIAQI
GE NAN

Assignees

合肥气象量子技术创新研究中心

Dates

Publication Date: 20260512
Application Date: 20260301

Claims (10)

1. The visibility inversion method for multi-parameter fusion of the light quantum radar is characterized by comprising the following steps of: Receiving a backward scattering echo photon signal through an atmospheric inclined path by a single photon avalanche diode array, and synchronously recording an arrival time stamp, a spatial position coordinate, polarization state information and wave front phase information of each photon; Constructing a distance gating photon counting sequence segmented along an inclined path based on the photon arrival time stamp, performing poisson distribution fitting on photon counting in each distance gate, and extracting the ratio of photon counting variance to mean value as an intensity flicker index; Calculating the instantaneous offset of the centroid of the light beam on the receiving plane based on the space position coordinates, carrying out power spectrum density analysis on the offset sequence in the continuous time window, and extracting the energy duty ratio of the main frequency band as the characteristic parameter of the light beam drift; Based on the polarization state information, reconstructing a polarization ellipse parameter of the echo photon by utilizing Stokes parameters, and calculating a polarization degree attenuation rate and a polarization azimuth angle rotation quantity which are used as sensitive indexes of aerosol particle shape and concentration distribution; Based on the wavefront phase information, reconstructing phase distortion caused by atmospheric turbulence by adopting a shack-Hartmann wavefront sensor, and calculating root mean square values of second to fifth order coefficients of the Zernike polynomials as quantitative characterization of atmospheric coherence length; the intensity flicker index, the beam drift characteristic parameter, the polarization degree attenuation rate, the polarization azimuth angle rotation amount and the Zernike low-order coefficient root mean square value are combined to form a multi-parameter fusion characteristic vector; Inputting the multi-parameter fusion feature vector into a pre-trained visibility inversion neural network model, wherein the visibility inversion neural network model is composed of an encoder-decoder structure, the encoder is a multi-head attention mechanism and time sequence convolution mixed network, the decoder is a fully-connected regression layer, and the visibility values corresponding to all distance gates on a slope path are output.
2. The method for inverting the visibility of multiparameter fusion of the light quantum radar according to claim 1, wherein the steps of constructing a distance gating photon counting sequence segmented along a diagonal path based on the photon arrival time stamp, performing poisson distribution fitting on photon counts in each distance gate, extracting the ratio of photon counting variance to mean value as an intensity scintillation index, and comprising the steps of: time interval based on laser pulse emission time Nanosecond dividing time windows; Accumulating the number of photons received by each range gate to form a photon counting sequence; for 100 consecutive laser pulse periods, the mean and variance of photon counts within each range gate are calculated and the ratio of variance to mean is defined as the intensity flicker index.
3. The method for inverting the visibility of the multiparameter fusion of the optical quantum radar according to claim 2, wherein calculating the instantaneous offset of the centroid of the light beam on the receiving plane based on the spatial position coordinates, performing power spectral density analysis on the offset sequence in the continuous time window, extracting the energy duty ratio of the main frequency band as the characteristic parameter of the light beam drift, comprises: Set the first The spatial coordinates of the detection units are At the moment of The number of received photons is The total number of effective detection units is Instantaneous shift of beam centroid in x-direction Instantaneous offset in the y-direction ; Carrying out power spectrum density analysis on the offset sequence by adopting a Welch method, wherein a window function is a Hanning window, the window length is 1 second, and the overlapping rate is 50%; and defining a main frequency band, and calculating the proportion of the power spectrum integral in the main frequency band to the total power spectrum integral as a beam drift characteristic parameter.
4. The visibility inversion method of claim 3 wherein reconstructing the polarization ellipse parameters of echo photons using stokes parameters based on said polarization state information, calculating the polarization degree decay rate and the polarization azimuth rotation amount comprises: Photon count intensity through horizontal polarization component Photon count intensity with perpendicular polarization component And +45 degree polarization component intensity And-45 degree polarization component intensity Right-hand circular polarization component intensity Intensity of left-hand circular polarization component Reconstructing Stokes parameters, total light intensity Intensity difference between horizontal and vertical polarization components Intensity difference of + -45 degree polarization component Difference in right-hand and left-hand circular polarization component intensities ; Calculating the attenuation rate of the polarization degree ; Calculating the polarization azimuth rotation amount 。
5. The method of claim 4, wherein reconstructing phase distortions caused by atmospheric turbulence using a shack-hartmann wavefront sensor based on the wavefront phase information, calculating root mean square values of second to fifth order coefficients of Zernike polynomials, comprising: the local wavefront slope output by the base Yu Xiake-Hartmann sensor is fitted with the first 20 terms of Zernike polynomial coefficients by a least square method; Extracting second to fifth order Zernike coefficients 、、、 Defocus, astigmatism and coma, respectively; Calculating root mean square value , Is the first The order Zernike coefficients.
6. The method for inverting the visibility of the multi-parameter fusion of the optical quantum radar according to claim 5, wherein the combination of the intensity flicker index, the beam drift characteristic parameter, the polarization degree attenuation rate, the polarization azimuth rotation amount, and the Zernike low-order coefficient root mean square value to form the multi-parameter fusion characteristic vector comprises the steps of: 5 characteristic parameters are arranged according to the sequence of the distance gates to form five-dimensional characteristic vectors; the whole inclined path forms a characteristic matrix with length of 10000 and dimension of 5 as a multi-parameter fusion characteristic vector.
7. The method for inverting the visibility of the multiparameter fusion of the light quantum radar according to claim 6, wherein the multiparameter fusion feature vector is input into a pre-trained visibility inversion neural network model, the visibility inversion neural network model is composed of an encoder-decoder structure, the encoder is a multi-head attention mechanism and time sequence convolution hybrid network, the decoder is a fully-connected regression layer, and the output is a visibility value corresponding to each distance gate on a diagonal path, and the method comprises the following steps: The encoder is composed of four layers of multi-head self-attention layers and time sequence convolution layers which are alternately stacked, the activation function is a linear function, and the output dimension is equal to the number of distance gates.
8. The method for performing visibility inversion of a multi-parameter fusion of a light quantum radar according to claim 7, wherein the training process of the visibility inversion neural network model comprises: collecting oblique path echo data of known visibility standard values under different meteorological conditions, extracting corresponding multi-parameter fusion feature vectors as input, and taking standard visibility values as supervision labels; performing end-to-end training by adopting a mean square error loss function; training was continued until the validation set loss was no drop for 10 consecutive rounds.
9. The method for inverting the visibility of the multi-parameter fusion of the light quantum radar according to claim 8, wherein the shack-Hartmann micro lens array is positioned on a receiving focal plane, and the micro lens units are in one-to-one correspondence with the single photon detection units and are used for sampling the local slope of the wavefront; Each microlens focuses the local wavefront to a four-quadrant photodiode below the microlens, and the wavefront slope of the local area is calculated by comparing the light intensity differences of the four quadrants, so that the whole wavefront phase distribution is reconstructed.
10. A visibility inversion system for multiparameter fusion of a light quantum radar, comprising: The single photon detection module is used for receiving echo photon signals back scattered by the atmospheric inclined path and synchronously outputting arrival time stamps, space position coordinates, polarization state information and wave front phase information of each photon; the intensity scintillation feature extraction module is used for constructing a distance gating photon counting sequence based on the photon arrival time stamp, calculating the ratio of the variance to the mean value of photon counting in each distance gate and generating an intensity scintillation index; The beam drift characteristic extraction module is used for calculating the instantaneous offset of the mass center of the beam based on the space position coordinates, carrying out power spectrum density analysis on the offset sequence, and extracting the energy duty ratio of the main frequency band as a beam drift characteristic parameter; The polarization evolution characteristic extraction module is used for reconstructing Stokes parameters based on the polarization state information and calculating the attenuation rate of the polarization degree and the rotation quantity of the polarization azimuth angle; The wave-front phase disturbance characteristic extraction module is used for reconstructing atmospheric turbulence phase distortion based on the wave-front phase information and calculating root mean square values of second to fifth order coefficients of the Zernike polynomials; The multi-parameter fusion module is used for combining the intensity flicker index, the beam drift characteristic parameter, the polarization degree attenuation rate, the polarization azimuth angle rotation amount and the Zernike low-order coefficient root mean square value into a multi-parameter fusion characteristic vector; The visibility inversion module is used for inputting the multi-parameter fusion feature vector into a pre-trained visibility inversion neural network model and outputting visibility values corresponding to each distance gate on the inclined path; The visibility inversion neural network model comprises an encoder and a decoder, wherein the encoder is formed by alternately stacking a multi-head self-attention layer and a time sequence convolution layer and is used for capturing nonlinear coupling relation and time sequence dynamic characteristics among multiple parameters, the decoder is formed by three full-connection layers, an activation function is a linear function, and the output dimension is equal to the number of distance gates.

Description

Visibility inversion method and system for multi-parameter fusion of light quantum radar Technical Field The invention belongs to the technical field of photoelectric detection and remote sensing, and particularly relates to a visibility inversion method and system for multi-parameter fusion of a light quantum radar. Background Along with the continuous improvement of the requirements of aviation, aerospace and high-precision meteorological observation on the detection of the visibility of the inclined plane atmosphere, the conventional visibility inversion method based on the light intensity attenuation model faces the problem. Such methods generally estimate the extinction coefficient by measuring the ratio of the intensities of the transmitted and received optical signals, assuming that the atmospheric medium is uniform and the transmission path is stable, and thus derive the visibility. In a real inclined path, atmospheric turbulence is ubiquitous, and the random fluctuation of the refractive index caused by the atmospheric turbulence can cause intensity flicker, phase distortion and spatial drift of a light beam, so that the monotonic corresponding relation between the light intensity and the path attenuation is seriously damaged. Under long distance, low elevation angle or high turbulence intensity conditions, the light intensity fluctuation can mask the real attenuation signal dominated by aerosol scattering, so that the visibility inversion result is deviated and even completely fails. Optical quantum radar is used as an emerging active remote sensing technology, and the unique property of a quantum light source is utilized to explore the detection capability beyond the classical limit. Compared with the traditional laser radar which relies on the intensity or phase information of a classical light field, the quantum radar can extract deep features of environmental disturbance through non-local association characteristics of entangled photon pairs. The time-frequency entangled photon pair presents strong quantum association in the time domain and the frequency domain, so that the time delay information transmitted by a coding path and the frequency mismatch information caused by medium disturbance can be coded simultaneously theoretically, and a physical basis is provided for multi-parameter joint perception. In the prior art, the synchronous, decoupled inversion of visibility and turbulence intensity is difficult to realize in an inclined Cheng Tuanliu environment whether a classical laser radar or a primarily explored quantum detection system. Conventional devices only collect single intensity or echo time information, cannot distinguish the mixed effect of aerosol attenuation and turbulent scattering, and even if partial polarization or coherence measurement is introduced, the capability of quantitatively characterizing quantum association degradation caused by turbulence is still lacking. There is no effective algorithm capable of separating photon time broadening parameters which independently reflect visibility and associated distortion parameters which characterize the intensity of path integral turbulence from coincidence count and frequency associated spectra of entangled photons. Under the inclined path detection scene with strong turbulence interference, a novel inversion frame integrating quantum optics and atmospheric physics is needed to break through the problems of parameter confusion, insufficient sensitivity, poor environmental adaptability and the like of the traditional method. Disclosure of Invention The invention provides a visibility inversion method and a visibility inversion system for multiparameter fusion of a light quantum radar, and aims to solve the technical problems that in visibility detection of a diagonal path such as aviation, aerospace and the like, light intensity flicker and light beam drift caused by atmospheric turbulence cause interference to visibility measurement based on intensity attenuation, and the conventional equipment cannot synchronously and accurately quantify the influence of turbulence. According to the invention, a multidimensional quantum optical observation system integrating photon arrival time distribution, photon counting statistical characteristics, polarization state evolution characteristics and wavefront phase disturbance information is constructed, and a coupled physical model of turbulence intensity and visibility is combined, so that high-precision and anti-interference inversion of the visibility of a diagonal path is realized. The invention provides a visibility inversion method for multi-parameter fusion of a light quantum radar, which comprises the following steps: Receiving a backward scattering echo photon signal through an atmospheric inclined path by a single photon avalanche diode array, and synchronously recording an arrival time stamp, a spatial position coordinate, polarization state information and wave front phase info