A method for measuring mirror seeing based on neural networks

Abstract

The present application relates to a kind of mirror seeing-fidelity measurement methods based on neural network, comprising the following steps: the multiple wavefront Zernike coefficients measured by the horizontal detection device are input with multiple temperature, the wavefront Zernike coefficient measured simultaneously by the vertical detection device is output, and neural network is constructed;The multiple wavefront Zernike coefficients measured by the horizontal detection device are input into neural network with multiple temperature, and mirror seeing-fidelity is obtained.The mirror seeing-fidelity measurement method based on neural network of the present application, by the measurement position of mirror seeing-fidelity from optical axis direction is transferred to horizontal direction, can realize the real-time measurement of mirror seeing-fidelity under the premise of not affecting telescope observation;Without long measuring device, and without telescope additional light splitting to measuring device, so as to help telescope to realize the effective detection of low contrast target.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of optical measurement technology, and in particular to a mirror seeing measurement method based on a neural network. BACKGROUND

[0002] The resolution capability of a telescope is proportional to the light collecting capability and its aperture, and as the demand for deep space exploration, dark sky exploration and exoplanet exploration increases, the aperture of the telescope is also gradually increasing. Currently, the aperture of an optical telescope has reached 30 meters, and the three 30-meter giant telescopes, GMT, TMT and E-ELT, are already under construction. However, as the aperture of the telescope continues to increase, the impact of seeing also becomes increasingly significant. Seeing is a physical quantity that evaluates the size of the impact of turbulence on target imaging, and mainly includes atmospheric seeing, dome seeing and mirror seeing. Poor seeing can cause imaging drift and diffusion in the back end, and also introduce additional errors for high-precision processing and detection applications. Atmospheric seeing and dome seeing have been well solved through long-term development. Compared with the first two types of seeing, mirror seeing has less impact on imaging, but its importance for imaging applications has gradually become prominent after the first two types of seeing have been well addressed.

[0003] Mirror seeing mainly refers to the deterioration of imaging and detection caused by mirror turbulence. Traditional mirror seeing measurement methods include temperature field method, interferometer method and scintillometer method. The temperature field method obtains the temperature distribution of the entire mirror by installing temperature sensors on the mirror of the telescope, and uses the relationship between temperature and refractive index to obtain the refractive index distribution of the mirror, thereby obtaining the mirror seeing. The interferometer method places an interferometer on the focal plane of the telescope, and uses the interference fringes to obtain the impact of mirror turbulence on the incident wavefront, thereby calculating the mirror seeing. The scintillometer method uses the fluctuation of photons to obtain the mirror seeing.

[0004] The temperature field method requires the installation of temperature sensors on the mirror of the telescope, and the higher the accuracy of the mirror seeing measurement to be achieved, the greater the number of temperature sensors that need to be installed, which puts higher requirements on the design of the mirror support system of the telescope. The interferometer method and the scintillometer method need to be co-optical with the observation equipment of the telescope and can only be measured during observation intervals, which not only affects the observation of the telescope itself, but also cannot achieve real-time measurement, making it difficult to guide subsequent related improvements. SUMMARY

[0005] The present application solves the technical problems in the prior art and achieves real-time mirror seeing measurement without affecting the observation of the telescope, and provides a mirror seeing measurement method based on a neural network.

[0006] To solve the above technical problems, the technical scheme of the present application is as follows:

[0007] A mirror seeing measurement method based on a neural network, the detection device suitable for the method comprising: a horizontal detection device and a vertical detection device; the horizontal detection device is used to measure multiple wavefront Zernike coefficients and multiple temperatures in the horizontal direction; the vertical detection device is used to measure wavefront Zernike coefficients in the vertical direction;

[0008] The method comprises the following steps:

[0009] The multiple wavefront Zernike coefficients and multiple temperatures measured by the horizontal detection device are taken as inputs, and the wavefront Zernike coefficients measured by the vertical detection device are taken as outputs to construct a neural network;

[0010] The multiple wavefront Zernike coefficients and multiple temperatures measured by the horizontal detection device are input into the neural network to obtain the mirror seeing.

[0011] In the above technical scheme, the horizontal detection device comprises: a wavefront sensor, a mirror, a holder, a temperature sensor, and a fiber laser;

[0012] The vertical detection device comprises: an interferometer;

[0013] The fiber laser is placed below the opening of the mirror, and a measurement light path is output from the fiber laser to realize turbulence measurement of the entire mirror;

[0014] The measurement light path output by the fiber laser is connected to the corresponding position on the holder, and the beam expansion and collimation of the measurement light are realized through the coupling device on the holder;

[0015] The wavefront sensor is placed to the left of the region to be measured of the mirror to receive the distorted wavefront after passing through the region to be measured, and the wavefront sensor needs to be aligned with the output end on the holder;

[0016] The temperature sensor is used to measure multiple temperature values on the mirror;

[0017] The interferometer is placed at a position where the focal point of the interferometer coincides with the focal point of the mirror, and the interferometer and the measurement light path in the horizontal direction jointly measure.

[0018] In the above technical scheme, the method comprises the following steps:

[0019] Step i: measuring the distortion of the turbulence to be measured above the mirror to the incident wavefront by using the wavefront sensor;

[0020] Step ii: measuring the influence of the to-be-measured turbulence on the vertical incident wavefront in the optical axis direction while measuring horizontally;

[0021] Step iii: constructing a neural network with the Zernike coefficients measured horizontally as input and the wavefront Zernike coefficients measured by the interferometer as output;

[0022] Step iv: removing the interferometer, inputting the Zernike coefficients measured horizontally and the temperature into the neural network to obtain the mirror seeing.

[0023] In the above technical solution, the number of measurement light paths output by the fiber laser is 12.

[0024] In the above technical solution, the number of temperature sensors matches the number of measurement paths, and the temperature sensors are placed below the mirror corresponding to each light path.

[0025] The present application has the following beneficial effects:

[0026] The mirror seeing measurement method based on a neural network of the present application can realize real-time measurement of mirror seeing without affecting the observation of the telescope by transferring the measurement position of mirror seeing from the optical axis direction to the horizontal direction. Meanwhile, placing the measurement light path in the horizontal direction does not require a long measurement device, and this non-common-path measurement method of the telescope does not require the telescope to additionally split light for the measurement device, thereby helping the telescope to effectively detect low-contrast targets. BRIEF DESCRIPTION OF DRAWINGS

[0027] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0028] Figure 1 It is a schematic diagram of the horizontal direction multi-path measurement of the mirror seeing measurement method based on a neural network of the present application.

[0029] Figure 2 It is a schematic diagram of the neural network training vertical and horizontal direction measurement devices.

[0030] Figure 3 It is a flowchart of the mirror seeing measurement.

[0031] The reference signs in the drawings are as follows:

[0032] 1-wavefront sensor; 2-mirror; 3-clamp; 4-temperature sensor; 5-fiber laser; 6-interferometer. DETAILED DESCRIPTION

[0033] The inventive concept of this invention is as follows: The method for measuring mirror seeing based on a neural network is grounded in the "Taylor frozen turbulence" assumption. It utilizes wavefront sensors to measure the influence of mirror turbulence along the path of the wavefront in the horizontal direction, while simultaneously increasing the number of measurement paths to obtain the influence of mirror turbulence on the incident wavefront in different regions of the entire mirror. After obtaining the influence of the entire mirror turbulence on multiple wavefronts in the horizontal direction, based on various isotropic assumptions and the inherent relationship between horizontal and vertical incident wavefronts, a neural network is constructed. This network uses the Zernike coefficients of the multi-path wavefront measurements as input and the Zernike coefficients of the wavefront measured by an interferometer in the vertical direction as output. Multiple temperature measurements are added to the input to constrain the results. In practical applications, only the Zernike coefficients and temperature values ​​of the wavefront in the horizontal direction need to be input into the neural network to obtain the influence of mirror turbulence on the incident wavefront of the vertical mirror in the optical axis direction.

[0034] This invention relates to a device suitable for detecting mirror seeing based on a neural network, mainly comprising a horizontal detection device and a vertical detection device. The horizontal detection device consists of a wavefront sensor, a mirror, a holder, a temperature sensor, and a fiber laser. The fiber laser is placed below the opening in the mirror, and its multi-channel measurement output is fixed to the holder, which is positioned slightly above the mirror. Wavefront sensors are placed at appropriate locations after each optical path passes through the turbulent flow being measured. The temperature sensor is placed on the back of the mirror below each optical path. The mirror seeing is measured by traversing the entire mirror surface through multiple optical paths. The vertical detection device consists of an interferometer. The interferometer is positioned so that the focal point of the mirror coincides with the focal point of the interferometer, and it performs measurements synchronously with the horizontal detection device, thereby generating target data for neural network training.

[0035] The present invention will now be described in detail with reference to the accompanying drawings.

[0036] The detection device to which the neural network-based method for measuring specular seeing is applicable is, for example... Figure 1 and Figure 2As shown, the main components include: a wavefront sensor 1, a mirror surface 2, a holder 3, a temperature sensor 4, and a fiber laser 5, a horizontal detection device, and an interferometer 6, a vertical detection device. The fiber laser 5 is placed below the opening of the mirror surface 2, and 12 light paths are output from the fiber laser 5 to measure the turbulence of the entire mirror surface 2. The 12 measurement light paths output by the fiber laser 5 are connected to the corresponding positions on the holder 3, and the coupling device on the holder 3 is used to expand and collimate the measurement light. The wavefront sensor 1 is placed to the left of the mirror surface 2 to be measured to receive the distorted wavefront after the measurement area, and the wavefront sensor 1 needs to be aligned with the output end of the holder 3. The interferometer 6 is placed at a position where the focal point coincides with the focal point of the mirror surface 2 to be measured, and the interferometer 6 and the 12 measurement light paths in the horizontal direction are used for measurement.

[0037] The principle of the mirror surface seeing measurement device based on the neural network is as shown in Figure 1 The 12 measurement light paths are output by the fiber laser 5, and the 12 measurement light paths are uniformly distributed in the entire measurement area of the mirror surface 2 to measure the entire mirror surface 2. The mirror surface seeing is generated due to the different refractive indexes at different positions caused by the different temperatures in the mirror surface area. The different refractive indexes at different positions cause the wavefront to be distorted after the light passes through the measurement area, and the wavefront sensor 1 is used to measure the distortion of the incident wavefront caused by the turbulence above the mirror surface 2 to evaluate the size of the mirror surface seeing. At the same time, based on the Taylor frozen turbulence assumption and the isotropic assumption, the wavefront measured horizontally is related to the wavefront measured vertically. The influence of the turbulence to be measured on the vertically incident wavefront is measured in the optical axis direction. The measured Zernike coefficients and the temperature of the multiple paths in the horizontal direction are used as inputs, and the Zernike coefficients measured by the interferometer 6 are used as outputs to construct a neural network. The training stage of the neural network needs to measure horizontally and vertically at the same time, and after the neural network is trained, the interferometer 6 can be removed, and only the horizontal measurement device is retained. In the actual scene, the Zernike coefficients and the temperature of the multiple measurement paths are input into the neural network to obtain the mirror surface seeing.

[0038] The number of temperature sensors 4 matches the number of measurement paths, and each temperature sensor 4 is placed below the mirror surface 2 corresponding to each light path. The position is determined according to the mirror surface 2 support system, and each temperature sensor 4 is fixed relative to each light path on the mirror surface 2. In the vertical direction, in addition to using the interferometer 6 to measure the wavefront, the wavefront sensor 1 can also be used to measure the wavefront.

[0039] The method for measuring the mirror surface seeing based on the neural network of the application is as shown in Figure 3 The method comprises the following steps:

[0040] Step i: measuring the distortion of the incident wavefront by the turbulence to be measured above the mirror 2 by using the wavefront sensor 1;

[0041] Step ii: measuring the influence of the turbulence to be measured on the vertical incident wavefront by measuring in the optical axis direction while measuring horizontally;

[0042] Step iii: constructing a neural network by taking the measured Zernike coefficients and the temperature in the horizontal direction as input, and the wavefront Zernike coefficients measured by the interferometer 6 as output;

[0043] Step iv: removing the interferometer 6, inputting the measured Zernike coefficients and the temperature in the horizontal direction into the neural network to obtain the mirror seeing.

[0044] The method for measuring the mirror seeing based on the neural network of the application is applied in the measurement, and the degree of the turbulence of the mirror to the distortion of the vertical incident wavefront is used as the evaluation method of the mirror seeing, which is mainly applied in the system integration and mirror processing of the telescope. In the selection of the measurement path, the actual application scene is selected, and the number of measurements can be increased to realize the precise measurement of the mirror seeing in the application scene with strong turbulence activity.

[0045] The method for measuring the mirror seeing based on the neural network of the application can realize the real-time measurement of the mirror seeing without affecting the observation of the telescope by transferring the measurement position of the mirror seeing from the optical axis direction to the horizontal direction. At the same time, the measurement path is placed in the horizontal direction without the need for a long measurement device, and the measurement method which is not co-path with the telescope does not need the telescope to additionally split light for the measurement device, thereby helping the telescope to realize the effective detection of the low-contrast target.

[0046] Obviously, the above embodiments are only examples for clearly illustrating, but not limit the embodiments. For those skilled in the art, on the basis of the above description, other different forms of changes or variations can also be made. Here, all the embodiments need not and cannot be exhausted. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A method of neural network based measurement of mirror seeing, characterized in that, The method is suitable for a detection device, which comprises a horizontal detection device and a vertical detection device; the horizontal detection device is used for measuring multiple wavefront Zernike coefficients and multiple temperatures in a horizontal direction; the vertical detection device is used for measuring wavefront Zernike coefficients in a vertical direction; The method comprises the following steps: Constructing a neural network by taking the multiple wavefront Zernike coefficients and multiple temperatures measured by the horizontal detection device as inputs and the wavefront Zernike coefficients measured by the vertical detection device as outputs; Taking the multiple wavefront Zernike coefficients and multiple temperatures measured by the horizontal detection device as inputs to the neural network to obtain the mirror seeing; The horizontal detection device comprises a wavefront sensor (1), a mirror (2), a holder (3), a temperature sensor (4) and a fiber laser (5); The vertical detection device comprises an interferometer (6); The fiber laser (5) is placed below an opening of the mirror (2) to output a measurement light path from the fiber laser (5) for realizing turbulence measurement of the whole mirror (2); The measurement light path output by the fiber laser (5) is connected to a corresponding position on the holder (3) and the measurement light is expanded and collimated through a coupling device on the holder (3); The wavefront sensor (1) is placed on the left of a region of the mirror (2) to be measured to receive a distorted wavefront after passing through the region to be measured, and the wavefront sensor (1) needs to be aligned with an output end on the holder (3); The temperature sensor (4) is used for measuring multiple temperature values on the mirror (2); The interferometer (6) is placed at a position where a focal point of the interferometer (6) coincides with a focal point of the mirror (2), and the interferometer (6) measures together with the measurement light path in the horizontal direction; The method specifically comprises the following steps: Step i: measuring the distortion of an incident wavefront caused by turbulence above the mirror (2) by using the wavefront sensor (1); Step ii: measuring the influence of the turbulence on a vertically incident wavefront in an optical axis direction while measuring in the horizontal direction; Step iii: constructing a neural network by taking the multiple wavefront Zernike coefficients and multiple temperatures measured in the horizontal direction as inputs and the wavefront Zernike coefficients measured by the interferometer (6) as outputs; Step iv: removing the interferometer (6) and inputting the multiple wavefront Zernike coefficients and multiple temperatures measured in the horizontal direction to the neural network to obtain the mirror seeing.

2. The method of neural network based specula seeing measurement according to claim 1, characterized in that, The number of measurement light paths output by the fiber laser (5) is 12.

3. The method of neural network based specula seeing measurement according to claim 1, characterized in that, The number of temperature sensors (4) matches the number of measurement paths, and each temperature sensor (4) is placed below the mirror (2) corresponding to each measurement path.

Images