Method and system for tuning optical fiber mode-locked laser based on image recognition algorithm

Abstract

The invention discloses a method and system for tuning optical fiber mode-locked laser based on an image recognition algorithm, and relates to the technical field of optical fiber laser mode locking. Converting the spectral image of the RGB color space into an HSV color space, and separating a spectral curve from an image background through color screening and filtering; establishing a full-connection neural network FNN classification identification model based on image linear transformation to carry out mode-locked and non-mode-locked state identification classification on the spectrum curve; determining mode locking intervals of different wavelengths through the PC angle interval corresponding to the visual mode locking state; traversing and searching in a mode locking interval of a specified wavelength, and stopping PC rotation when a spectral image similar to the target image is detected. Therefore, mode locking states under two different wave bands are intelligently tuned.

Description

Technical Field

[0001] The invention relates to the technical field of optical fiber laser mode locking, and in particular to a method and a system for tuning an optical fiber mode locking laser based on an image recognition algorithm. Background Art

[0002] The development and needs of fiber-optic communication technology have promoted the development of multi-wavelength fiber lasers. This type of laser has the advantages of a large number of output wavelengths, high beam quality, high conversion efficiency and compact structure. In addition, in the fields of time imaging and microscopy, Raman scattering spectroscopy, nonlinear frequency conversion, multi-color pump-probe spectroscopy, mid-infrared difference frequency generation and differential radar, etc., stable and tight synchronization of light pulses of different wavelengths is required. Therefore, dual-wavelength fiber lasers have received widespread attention and have been deeply studied and developed.

[0003] Nonlinear polarization rotation (NPR) technology is a fiber laser structure used to generate and maintain soliton pulses. It is the first and most widely used structure of soliton pulse fiber lasers. Since NPR fiber lasers have great flexibility in wavelength and gain tuning range, NPR technology is also one of the important methods to realize dual-wavelength fiber lasers. Compared with other mode-locking technologies, NPR fiber lasers have a simple structure and only contain two polarization controllers (PCs). They do not require complex components such as external modulators or gratings, making system design and manufacturing easier. However, the study of dual wavelengths in NPR fiber lasers is still an emerging field. Under traditional conditions, the mode-locking state in different bands only appears under certain PC angle combinations. The process of manually adjusting the mode-locking in different bands using NPR technology is very random and accidental, and the process may take up to several hours. In addition, after the fiber laser loses lock due to external environmental influences such as ambient temperature and vibration, it takes a lot of time to readjust.

[0004] Intelligent mode-locking methods based on machine learning or deep learning have become a hot topic in recent years. For example, K-means and LSTM have been successfully applied to the automatic mode-locking of traditional soliton fiber lasers. However, currently, the numerical input of the spectral curve is used, the training takes a long time, and they are all for single-wavelength traditional soliton mode-locking. There is no intelligent mode-locking method with adjustable dual wavelengths. Summary of the invention

[0005] In order to solve the deficiencies mentioned in the above background technology, the object of the present invention is to provide a method and system for tuning fiber mode-locked laser based on image recognition algorithm.

[0006] In a first aspect, the purpose of the present invention can be achieved by the following technical solution: a method for tuning a fiber mode-locked laser based on an image recognition algorithm, the method comprising the following steps:

[0007] Receiving a spectrometer image, preprocessing the spectrometer image to obtain a preprocessed spectrometer image, generating an image data set from the preprocessed spectrometer image, and performing a labeling operation on the image data set to obtain a labeled image data set, wherein the spectrometer image is collected based on a central wavelength set by the spectrometer;

[0008] Input the labeled image data set into a pre-established FNN-based classification and recognition model for training, obtain the trained FNN-based classification and recognition model, obtain the spectrometer image to be identified, input the spectrometer image to be identified into the trained FNN-based classification and recognition model, and output the label of the image;

[0009] Based on the image label and the central wavelength set by the spectrometer, a scatter plot is used to draw the PC rotation angle combination of all classification labels for the mode-locked image. The data corresponding to different central wavelengths are marked with different colors, and the PC angle interval corresponding to each central wavelength where the mode-locked state occurs is obtained as the mode-locking interval;

[0010] Specify the central wavelength, select the corresponding mode-locking interval according to the specified central wavelength, and control the servo through the program to traverse the PC rotation angle combination in an orderly manner according to the fixed step length. When the target image appears, stop the PC rotation to achieve the tuning of the fiber mode-locking.

[0011] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the acquisition process of the spectrometer image is as follows:

[0012] The central wavelength of the spectrometer is set to be close to the central wavelengths of the two solitons generated by the laser cavity, and the servo is controlled by the program to rotate the PC in an orderly manner according to a fixed step size, and a spectrometer image is collected each time it rotates.

[0013] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the process of preprocessing the spectrometer image:

[0014] Input the image and convert the spectrum of the RGB color space into the HSV color space. The conversion formula is:

[0015] V = mmax(R, G, B)

[0016]

[0017] In the formula, R, G, and B are the values ​​of the three color components of an image, namely red, green, and blue, respectively, and H, S, and V are the values ​​of the three components of an image, namely hue, saturation, and brightness, respectively. By setting upper and lower thresholds for the H, S, and V parameters and using mask extraction, the image containing only the spectral curve can be separated from the converted image;

[0018] By setting the rectangular convolution kernel and the morphological operation of the opening operation, the separated image is filtered to remove noise;

[0019] Use Crop center cropping to get a square image in the center of the image;

[0020] Convert the obtained image into grayscale image for output.

[0021] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: the FNN based on the FNN classification and recognition model is a fully connected neural network based on image linear transformation, and the fully connected neural network based on image linear transformation includes: an input layer, a linear layer, four hidden layers, a softmax layer and an output layer.

[0022] In combination with the first aspect, in some implementations of the first aspect, the method further includes: a process of inputting the labeled image data set into a pre-established FNN-based classification and recognition model for training:

[0023] Divide the labeled image dataset into training set and test set;

[0024] The input layer of the fully connected neural network based on image linear transformation receives the image data represented by the input two-dimensional matrix, and the linear layer converts the two-dimensional matrix into a one-dimensional array and inputs it into the hidden layer; each layer of neurons adopts a fully connected mode, and the activation function is the relu function. The mapping relationship of the function is as follows:

[0025]

[0026] Among them, x is the input of each neuron, and f(x) is the output after passing through the neuron;

[0027] The last layer activation function is the tanh function, and the mapping relationship of the function is as follows:

[0028]

[0029] Where x is the input of each neuron, f(x) is the output after passing through the neuron; sinh(x) is the hyperbolic sine function, cosh(x) is the hyperbolic cosine function, and tanh(x) is the hyperbolic tangent function;

[0030] The output layer uses the softmax function to convert the output result of the previous neural network into a probability value P between 0 and 1. i , the formula is:

[0031]

[0032] Among them, z is a vector, which is the output result of the previous layer of neural network, z i and z j is one of the elements;

[0033] Set the loss function and optimizer. The loss function uses the cross entropy loss function. The optimizer uses the Adam algorithm.

[0034] Use the training set to train the model, calculate the training loss after each training, and test it on the test set. If the training loss value converges and the average test accuracy is greater than the set threshold, stop the model training and output the trained FNN-based classification and recognition model.

[0035] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the process of controlling the servo to traverse the PC rotation angle combination in order according to a fixed step length by program:

[0036] Initialize the servo angle and restore it to the vertical position;

[0037] Determine the required central wavelength and select the corresponding mode-locking interval as the input of the ergodic control algorithm;

[0038] The ergodic control algorithm is used to control the servo to traverse the PC angle combination, with a change step of 1° to 1.5°. Each time the PC angle is changed, the output image of the spectrometer is detected in real time.

[0039] If it is detected that the output image is similar to the target image, the program ends;

[0040] If it is detected that the output image is not similar to the target image, the detection continues until the program ends.

[0041] In a second aspect, in order to achieve the above-mentioned object, the present invention discloses a system for tuning a fiber mode-locked laser based on an image recognition algorithm, comprising:

[0042] An image processing module is used to receive a spectrometer image, preprocess the spectrometer image to obtain a preprocessed spectrometer image, generate an image data set from the preprocessed spectrometer image, and perform a labeling operation on the image data set to obtain a labeled image data set, wherein the spectrometer image is collected based on a central wavelength set by the spectrometer;

[0043] A model training module is used to input the labeled image data set into a pre-established FNN-based classification and recognition model for training, obtain the trained FNN-based classification and recognition model, obtain the spectrometer image to be identified, input the spectrometer image to be identified into the trained FNN-based classification and recognition model, and output the label of the image;

[0044] The interval determination module is used to draw the PC rotation angle combination of all classification labels for mode-locked images based on the image label and the central wavelength set by the spectrometer using a scatter plot, and the data corresponding to different central wavelengths are marked with different colors to obtain the PC angle interval corresponding to each central wavelength where the mode-locked state occurs as the mode-locking interval;

[0045] The mode-locking tuning module specifies the center wavelength, selects the corresponding mode-locking interval according to the specified center wavelength, and controls the servo to traverse the PC rotation angle combination in an orderly manner according to a fixed step size through the program. When the target image appears, the PC rotation is stopped to achieve the tuning of the fiber mode-locking.

[0046] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a terminal device is disclosed, including a memory, a processor, and a computer program stored in the memory and capable of running on the processor, wherein the memory stores a computer program capable of running on the processor, and when the processor loads and executes the computer program, a method for tuning a fiber mode-locked laser based on an image recognition algorithm as described above is adopted.

[0047] In another aspect of the present invention, in order to achieve the above-mentioned purpose, a computer-readable storage medium is disclosed, in which a computer program is stored. When the computer program is loaded and executed by a processor, a method for tuning a fiber mode-locked laser based on an image recognition algorithm as described above is adopted.

[0048] Beneficial effects of the present invention:

[0049] Based on the difference of spectral images under mode-locked and non-mode-locked states, the present invention constructs an identification and classification model based on linear transformation, and innovatively searches for all traditional soliton mode-locked intervals through image recognition methods. Based on this, soliton mode-locked images in two different bands can be automatically adjusted as needed.

[0050] When the present invention realizes soliton mode locking at different central wavelength bands, the tunable range of the central wavelength can be determined, that is, the interval covering all central wavelengths.

[0051] The present invention uses common rotary servos, servo control boards, Raspberry Pi, and thermoplastics to build a mechanical module for controlling a PC, thereby realizing electric control of the PC at a very low cost and compensating for the problems of instability, randomness, low precision, and the like in manual adjustment of the PC. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0053] Figure 1 It is a schematic flow chart of the method of the present invention;

[0054] Figure 2 It is a schematic diagram of a tunable dual-wavelength intelligent mode locking system based on spectral image recognition provided by an embodiment of the present invention;

[0055] Figure 3 It is a schematic diagram of the image processing flow of the method of the present invention;

[0056] Figure 4 It is a neural network architecture diagram for image classification and recognition according to an embodiment of the present invention;

[0057] Figure 5 It is a principle diagram of the wavelength tuning algorithm involved in the method of the present invention;

[0058] Figure 6 is a diagram of experimental results of dual wavelength tuning in an embodiment of the present invention;

[0059] Figure 7 This is an example diagram of a traditional soliton boundary that can be formed near 1530nm in a laser cavity in an embodiment of the present invention;

[0060] Figure 8 This is an example diagram of a traditional soliton boundary that can be formed near 1560nm in a laser cavity in an embodiment of the present invention;

[0061] Fig. 9 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Embodiment 1:

[0064] The following is an introduction to the relevant terms involved in the embodiments of the present application:

[0065] The mode-locked fiber laser based on the NPR structure is a passive mode-locked fiber laser. Passive mode locking refers to the use of the saturable absorption characteristics of passive devices to modulate the laser to obtain pulses. Saturable absorbers are divided into natural saturable absorbers and equivalent saturable absorbers. Natural saturable absorbers are commonly used semiconductor saturable absorber mirrors, or are made of materials such as carbon nanotubes and graphene. The NPR technology uses a polarization controller PC and a polarization-dependent isolator to construct an equivalent saturable absorber, adjust the self-phase modulation and cross-phase modulation of the pulse in the light, and thus achieve mode locking to obtain ultrashort pulses. Figure 1 As shown, a method for tuning a fiber mode-locked laser based on an image recognition algorithm comprises the following steps:

[0066] S101: receiving a spectrometer image, preprocessing the spectrometer image to obtain a preprocessed spectrometer image, generating an image data set from the preprocessed spectrometer image, and performing a labeling operation on the image data set to obtain a labeled image data set, wherein the spectrometer image is collected based on a central wavelength set by the spectrometer;

[0067] Specifically, Figure 2 As shown in the figure, a hardware system relied on by a tunable dual-wavelength intelligent mode locking method based on spectral image recognition includes three parts: a soliton fiber laser based on NPR structure and erbium-doped fiber, a steering gear control module and a spectrometer; in the soliton fiber laser, the erbium-doped fiber (EDF) is 2.6m to 4m long, the single-mode fiber (SMF) is 15m to 20m long, and the pump (LD) power is 160mW to 200mW, which can realize traditional soliton mode locking near 1530nm and 1560nm; the steering gear control module includes a rotary steering gear, a mechanical structure made of thermoplastic plastic, a control panel composed of a steering gear control board, a Raspberry Pi and a computer, and the Raspberry Pi and the computer are connected by WIFI.

[0068] The acquisition process of the spectrometer image is as follows:

[0069] The center wavelength of the spectrometer is set to be close to the center wavelength of the two solitons generated by the laser cavity, and the servo is controlled by the program to rotate the PC in an orderly manner according to a fixed step length, and a spectrometer image is collected every rotation;

[0070] The process of preprocessing the spectrometer image:

[0071] like Figure 3 As shown, input the image and convert the spectrum of the RGB color space into the HSV color space. The conversion formula is:

[0072] V = mmax(R, G, B)

[0073]

[0074] In the formula, R, G, and B are the values ​​of the three color components of an image, namely red, green, and blue, respectively, and H, S, and V are the values ​​of the three components of an image, namely hue, saturation, and brightness, respectively. By setting upper and lower thresholds for the H, S, and V parameters and using mask extraction, the image containing only the spectral curve can be separated from the converted image;

[0075] By setting the rectangular convolution kernel and the morphological operation of the opening operation, the separated image is filtered to remove noise;

[0076] Use Crop center cropping to get a square image in the center of the image;

[0077] Convert the obtained image into grayscale image for output.

[0078] S102: input the labeled image data set into a pre-established FNN-based classification and recognition model for training, obtain the trained FNN-based classification and recognition model, obtain the spectrometer image to be recognized, input the spectrometer image to be recognized into the trained FNN-based classification and recognition model, and output the label of the image;

[0079] The FNN based on the FNN classification and recognition model is a fully connected neural network based on image linear transformation, such as Figure 4 As shown, the fully connected neural network based on image linear transformation includes: an input layer, a linear layer, four hidden layers, a softmax layer and an output layer.

[0080] The process of inputting the labeled image dataset into the pre-established FNN-based classification and recognition model for training:

[0081] Divide the labeled image dataset into training set and test set;

[0082] The input layer of the fully connected neural network based on image linear transformation receives the image data represented by the input two-dimensional matrix, and the linear layer converts the two-dimensional matrix into a one-dimensional array and inputs it into the hidden layer; each layer of neurons adopts a fully connected mode, and the activation function is the relu function. The mapping relationship of the function is as follows:

[0083]

[0084] Among them, x is the input of each neuron, and f(x) is the output after passing through the neuron;

[0085] The last layer activation function is the tanh function, and the mapping relationship of the function is as follows:

[0086]

[0087] Where x is the input of each neuron, f(x) is the output after passing through the neuron; sinh(x) is the hyperbolic sine function, cosh(x) is the hyperbolic cosine function, and tanh(x) is the hyperbolic tangent function;

[0088] The output layer uses the softmax function to convert the output result of the previous neural network into a probability value P between 0 and 1. i , the formula is:

[0089]

[0090] Among them, z is a vector, which is the output result of the previous layer of neural network, z i and z j is one of the elements; in this embodiment, the classification recognition model outputs only two results, so the z vector contains 2 elements.

[0091] Set the loss function and optimizer. The loss function uses the cross entropy loss function. The optimizer uses the Adam algorithm.

[0092] The training set is used for model training, and the number of samples used in each iteration is set to 10 to 20. The training loss is calculated after each training, and the test is performed on the test set. If the training loss value converges and the average test accuracy is greater than the set threshold of 95%, the model training is stopped and the trained FNN-based classification recognition model is output.

[0093] S103: Based on the image label and the central wavelength set by the spectrometer, a scatter plot is used to draw the PC rotation angle combination of all the classification labels as the mode-locking images, and the data corresponding to different central wavelengths are marked with different colors, and the PC angle interval corresponding to each central wavelength where the mode-locking state occurs is obtained as the mode-locking interval;

[0094] S104: specifying a central wavelength, selecting a corresponding mode-locking interval according to the specified central wavelength, and controlling the servo to traverse the PC rotation angle combination in an orderly manner according to a fixed step length through a program, and stopping the PC rotation when a target image appears, thereby realizing tuning of the optical fiber mode-locking.

[0095] like Figure 5 As shown in the figure, the process of controlling the servo to traverse the PC rotation angle combination in order according to the fixed step length is:

[0096] Initialize the servo angle and restore it to the vertical position;

[0097] Determine the required central wavelength and select the corresponding mode-locking interval as the input of the ergodic control algorithm;

[0098] The ergodic control algorithm is used to control the servo to traverse the PC angle combination, with a change step of 1° to 1.5°. Each time the PC angle is changed, the output image of the spectrometer is detected in real time.

[0099] If it is detected that the output image is similar to the target image, the program ends;

[0100] If it is detected that the output image is not similar to the target image, the detection continues until the program ends.

[0101] Specifically, the present invention is further described below by way of embodiments:

[0102] In order to verify the effectiveness of the method proposed in Example 1, this method was used in this example to successfully achieve tunable dual-wavelength intelligent mode locking and reproduce the single soliton mode locking spectra of two bands. Figure 6 As shown in Figure 2, when the required central wavelength is specified to be around 1530nm, the experimental results are as follows: Figure 6 The middle grey curve shows that the central wavelength is 1530.9nm. When the required central wavelength is specified to be around 1560nm, the experimental results are as follows Figure 6 As shown by the black curve in the middle, the central wavelength is 1562.2nm.

[0103] The present invention focuses on the difference between the spectral images in the mode-locked and non-mode-locked states, constructs a recognition and classification model based on a fully connected neural network with linear transformation, and accurately realizes the recognition of the mode-locked state in the fiber laser cavity. In this process, the RGB spectral image is converted to HSV color space, and the spectral curve image is extracted in combination with a filtering algorithm, which greatly reduces the background and noise interference of the image to be identified and improves the training and operation speed of the model. The specific training time of the model is as fast as 49.65 seconds, which greatly shortens the time consumption compared to the traditional simulation method that takes several hours.

[0104] At the same time, since this method defines the mode-locking interval of each wavelength, the soliton mode-locking images of different wavelengths can be adjusted in the interval by traversing the control program, so the central wavelength of all soliton mode-locking can be calculated to obtain the central wavelength range of mode-locking in each band. Figure 7 As shown, the central wavelength boundaries of the locked mode in the 1530nm band are 1529.4nm and 1531.9nm; Figure 8 As shown, the central wavelength boundaries of mode locking in the 1560nm band are 1560.0nm and 1563.3nm. In this embodiment, in these two ranges, the PC angle can be fine-tuned by the ergodic control algorithm to tune the mode locking state of different central wavelengths.

[0105] Embodiment 2: In the second aspect, as Fig. 9 As shown, in order to achieve the above-mentioned purpose, the present invention discloses a system for tuning a fiber mode-locked laser based on an image recognition algorithm, comprising:

[0106] An image processing module 11 is used to receive a spectrometer image, preprocess the spectrometer image to obtain a preprocessed spectrometer image, generate an image data set from the preprocessed spectrometer image, and perform a labeling operation on the image data set to obtain a labeled image data set, wherein the spectrometer image is collected based on a central wavelength set by the spectrometer;

[0107] The model training module 12 is used to input the labeled image data set into a pre-established FNN-based classification and recognition model for training, obtain the trained FNN-based classification and recognition model, obtain the spectrometer image to be identified, input the spectrometer image to be identified into the trained FNN-based classification and recognition model, and output the label of the image;

[0108] The interval determination module 13 is used to draw the PC rotation angle combination of all the classification labels as the mode-locking images based on the image label and the central wavelength set by the spectrometer using a scatter plot, and the data corresponding to different central wavelengths are marked with different colors, and the PC angle interval corresponding to each central wavelength where the mode-locking state occurs is obtained as the mode-locking interval;

[0109] The mode-locking tuning module 14 specifies the central wavelength, selects the corresponding mode-locking interval according to the specified central wavelength, and controls the servo to traverse the PC rotation angle combination in an orderly manner according to a fixed step length through a program. When the target image appears, the PC rotation is stopped to achieve the tuning of the optical fiber mode-locking.

[0110] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0111] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0112] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0113] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. A method for tuning a fiber mode-locked laser based on an image recognition algorithm, characterized in that: The method comprises the following steps: Receiving a spectrometer image, preprocessing the spectrometer image to obtain a preprocessed spectrometer image, generating an image data set from the preprocessed spectrometer image, and performing a labeling operation on the image data set to obtain a labeled image data set, wherein the spectrometer image is collected based on a central wavelength set by the spectrometer; Input the labeled image data set into a pre-established FNN-based classification and recognition model for training, obtain the trained FNN-based classification and recognition model, obtain the spectrometer image to be identified, input the spectrometer image to be identified into the trained FNN-based classification and recognition model, and output the label of the image; Based on the image label and the central wavelength set by the spectrometer, a scatter plot is used to draw the PC rotation angle combination of all classification labels for the mode-locked image. The data corresponding to different central wavelengths are marked with different colors, and the PC angle interval corresponding to each central wavelength where the mode-locked state occurs is obtained as the mode-locking interval; Specify the central wavelength, select the corresponding mode-locking interval according to the specified central wavelength, and control the servo through the program to traverse the PC rotation angle combination in an orderly manner according to the fixed step length. When the target image appears, stop the PC rotation to achieve the tuning of the fiber mode-locking.

2. The method for tuning a fiber mode-locked laser based on an image recognition algorithm according to claim 1, characterized in that: The acquisition process of the spectrometer image is as follows: The central wavelength of the spectrometer is set to be close to the central wavelengths of the two solitons generated by the laser cavity, and the servo is controlled by the program to rotate the PC in an orderly manner according to a fixed step size, and a spectrometer image is collected each time it rotates.

3. The method for tuning a fiber mode-locked laser based on an image recognition algorithm according to claim 2, characterized in that: The process of preprocessing the spectrometer image: Input the image and convert the spectrum of the RGB color space into the HSV color space. The conversion formula is: V = mmax(R, G, B) In the formula, R, G, and B are the values ​​of the three color components of an image, namely red, green, and blue, respectively, and H, S, and V are the values ​​of the three components of an image, namely hue, saturation, and brightness, respectively. By setting upper and lower thresholds for the H, S, and V parameters and using mask extraction, the image containing only the spectral curve can be separated from the converted image; By setting the rectangular convolution kernel and the morphological operation of the opening operation, the separated image is filtered to remove noise; Use Crop center cropping to get a square image in the center of the image; Convert the obtained image into grayscale image for output.

4. The method for tuning a fiber mode-locked laser based on an image recognition algorithm according to claim 1, characterized in that: The FNN based on the FNN classification and recognition model is a fully connected neural network based on image linear transformation, and the fully connected neural network based on image linear transformation includes: an input layer, a linear layer, four hidden layers, a softmax layer and an output layer.

5. The method for tuning a fiber mode-locked laser based on an image recognition algorithm according to claim 1, characterized in that: The process of inputting the labeled image data set into the pre-established FNN-based classification and recognition model for training: Divide the labeled image dataset into training set and test set; The input layer of the fully connected neural network based on image linear transformation receives the image data represented by the input two-dimensional matrix, and the linear layer converts the two-dimensional matrix into a one-dimensional array and inputs it into the hidden layer; each layer of neurons adopts a fully connected mode, and the activation function is the relu function. The mapping relationship of the function is as follows: Among them, x is the input of each neuron, and f(x) is the output after passing through the neuron; The last layer activation function is the tanh function, and the mapping relationship of the function is as follows: Where x is the input of each neuron, f(x) is the output after passing through the neuron; sinh(x) is the hyperbolic sine function, cosh(x) is the hyperbolic cosine function, and tanh(x) is the hyperbolic tangent function; The output layer uses the softmax function to convert the output result of the previous neural network into a probability value P between 0 and 1. i , the formula is: Among them, z is a vector, which is the output result of the previous layer of neural network, z i and z j is one of the elements; Set the loss function and optimizer. The loss function uses the cross entropy loss function. The optimizer uses the Adam algorithm. Use the training set to train the model, calculate the training loss after each training, and test it on the test set. If the training loss value converges and the average test accuracy is greater than the set threshold, stop the model training and output the trained FNN-based classification and recognition model.

6. The method for tuning a fiber mode-locked laser based on an image recognition algorithm according to claim 1, characterized in that: The process of controlling the servo by program to traverse the PC rotation angle combination in order according to the fixed step length: Initialize the servo angle and restore it to the vertical position; Determine the required central wavelength and select the corresponding mode-locking interval as the input of the ergodic control algorithm; The ergodic control algorithm is used to control the servo to traverse the PC angle combination, with a change step of 1° to 1.5°. Each time the PC angle is changed, the output image of the spectrometer is detected in real time. If it is detected that the output image is similar to the target image, the program ends; If it is detected that the output image is not similar to the target image, the detection continues until the program ends.

7. A system for tuning fiber mode-locked laser based on image recognition algorithm, characterized in that: include: An image processing module is used to receive a spectrometer image, preprocess the spectrometer image to obtain a preprocessed spectrometer image, generate an image data set from the preprocessed spectrometer image, and perform a labeling operation on the image data set to obtain a labeled image data set, wherein the spectrometer image is collected based on a central wavelength set by the spectrometer; A model training module is used to input the labeled image data set into a pre-established FNN-based classification and recognition model for training, obtain the trained FNN-based classification and recognition model, obtain the spectrometer image to be identified, input the spectrometer image to be identified into the trained FNN-based classification and recognition model, and output the label of the image; The interval determination module is used to draw the PC rotation angle combination of all classification labels for mode-locked images based on the image label and the central wavelength set by the spectrometer using a scatter plot, and the data corresponding to different central wavelengths are marked with different colors to obtain the PC angle interval corresponding to each central wavelength where the mode-locked state occurs as the mode-locking interval; The mode-locking tuning module specifies the center wavelength, selects the corresponding mode-locking interval according to the specified center wavelength, and controls the servo to traverse the PC rotation angle combination in an orderly manner according to a fixed step size through the program. When the target image appears, the PC rotation is stopped to achieve the tuning of the fiber mode-locking.

8. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and capable of running on the processor, characterized in that: The memory stores a computer program that can be run on the processor. When the processor loads and executes the computer program, a method for tuning a fiber mode-locked laser based on an image recognition algorithm according to any one of claims 1 to 6 is adopted.

9. A computer-readable storage medium having a computer program stored therein, characterized in that: When the computer program is loaded and executed by the processor, a method for tuning a fiber mode-locked laser based on an image recognition algorithm as described in any one of claims 6 to 7 is adopted.

Images