CN116086497A - Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network - Google Patents

Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network Download PDF

Info

Publication number
CN116086497A
CN116086497A CN202310380651.7A CN202310380651A CN116086497A CN 116086497 A CN116086497 A CN 116086497A CN 202310380651 A CN202310380651 A CN 202310380651A CN 116086497 A CN116086497 A CN 116086497A
Authority
CN
China
Prior art keywords
neural network
sub
output signal
training
loss function
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202310380651.7A
Other languages
Chinese (zh)
Other versions
CN116086497B (en
Inventor
王建伟
宋晓林
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Polytechnic Navigation Control Technology Co ltd
Original Assignee
Beijing Polytechnic Navigation Control Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Polytechnic Navigation Control Technology Co ltd filed Critical Beijing Polytechnic Navigation Control Technology Co ltd
Priority to CN202310380651.7A priority Critical patent/CN116086497B/en
Publication of CN116086497A publication Critical patent/CN116086497A/en
Application granted granted Critical
Publication of CN116086497B publication Critical patent/CN116086497B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/761—Proximity, similarity or dissimilarity measures
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
    • G01C25/00—Manufacturing, calibrating, cleaning, or repairing instruments or devices referred to in the other groups of this subclass
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Multimedia (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Molecular Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Manufacturing & Machinery (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Gyroscopes (AREA)

Abstract

The invention provides an output correction method, a device and a storage medium under a low angular rate of a fiber optic gyroscope based on a neural network, wherein the method comprises the following steps: the method comprises the following steps of S101, collecting a current infrared image, working temperature and an output signal of a fiber optic gyroscope, wherein the output signal is generated by the fiber optic gyroscope at a low angular rate; a prediction step S102, namely inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, wherein the neural network performs training based on an optimized loss function; a correction step S103, processing based on the output signal and the first predicted output signal to obtain a corrected output signal; the neural network comprises two sub-neural networks, and the two sub-neural networks are jointly trained based on the optimized loss function, so that the overall prediction performance of the two sub-neural networks is optimal, and the accuracy of output correction is improved.

Description

Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network
Technical Field
The invention relates to the technical field of artificial intelligence and optical fiber gyroscope calibration, in particular to an output correction method and device based on a neural network for an optical fiber gyroscope at a low angular rate and a storage medium.
Background
Chinese patent application CN201811530095.2 discloses an improved method for scale factor nonlinearity of optical fiber gyroscope at low angular rate, in each working cycle of optical fiber gyroscope, the lowest bit of data collected by a/D converter is stored and combined into a binary number. And setting a register in the FPGA by taking the binary number as a basic sequence, and generating an inverse repeated M sequence through the register. And the periodic electronic crosstalk error caused by resetting is prevented by the modulation mode of the inverse repeated M sequence. Because the inverse repeated M sequence realizes real white noise relative to the M sequence and the mean value bit is zero, the error caused by other noise can be greatly reduced while the periodic electronic crosstalk is eliminated. The method adopts the collected analog data to carry out analog-to-digital conversion and then carries out processing, so that the improvement performance is limited, the output of the optical fiber gyroscope under the low angular rate can not be corrected, or even if the output is corrected, the reliability and the accuracy are poor.
The neural network is generally used for prediction, for example, after training the neural network according to the sample set data, prediction is performed based on the acquired data, which is a basic principle of the neural network, but how to perform prediction correction on the output of the optical fiber gyroscope under the low angular rate based on the neural network, that is, how to improve the accuracy and the reliability of correction based on the neural network is a technical problem.
In the prior art, the predicted value is generally directly used as the output signal after calibration, and this way is not applicable to output correction of the optical fiber gyroscope at a low angular rate, because the predicted value may be greatly different from the actual output value, and the reliability of the predicted value is low.
Disclosure of Invention
The present invention proposes the following technical solution to one or more of the above technical drawbacks of the prior art.
An output correction method of a fiber optic gyroscope based on a neural network under a low angular rate is characterized by comprising the following steps:
the method comprises the steps of collecting current infrared images, working temperature and output signals of the optical fiber gyroscope, wherein the output signals are generated by the optical fiber gyroscope at a low angular rate;
a prediction step, namely inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, wherein the neural network performs training based on an optimized loss function;
a correction step of processing based on the output signal and the first predicted output signal to obtain a corrected output signal;
the neural network comprises a first sub-neural network and a second sub-neural network, the first sub-neural network is trained based on sample infrared images in a training sample set, the second sub-neural network is trained based on sample working temperatures and sample output signals in the training sample set, and the first sub-neural network uses an loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function, wherein the optimized loss functionLThe method comprises the following steps:
Figure SMS_1
;
Figure SMS_2
;
Figure SMS_3
;
wherein ,
Figure SMS_4
representing the first of a training sample setiAn infrared image of the individual training samples is obtained,/>
Figure SMS_5
representation trainingTraining sample setiWorking temperature of each training sample, +.>
Figure SMS_6
Representing the first of a training sample setiOutput signal of training sample, +.>
Figure SMS_7
Representing a supervision value->
Figure SMS_8
Respectively represent predicted values,/->
Figure SMS_9
Respectively represent the corresponding weight values.
Still further, filtering the acquired infrared image based on the operating temperature to discard unsuitable infrared images.
Still further, the filtering operation includes: and acquiring a standard infrared image corresponding to the working temperature from a database, calculating the similarity between the acquired infrared image and the standard infrared image, discarding the acquired infrared image if the similarity is smaller than a first threshold value, continuing to acquire the infrared image, and judging the acquired infrared image to be qualified if the similarity is larger than or equal to the first threshold value.
Further, determining
Figure SMS_10
The method comprises the following steps:
Figure SMS_11
；
Figure SMS_12
；
wherein ,
Figure SMS_13
representing the maximum value of the loss function of the first sub-neural network during training, < >>
Figure SMS_14
Representing the minimum value of the loss function of the first sub-neural network during training, +.>
Figure SMS_15
Representing the maximum value of the loss function of the second sub-neural network during training, < >>
Figure SMS_16
Representing the minimum of the loss function of the second sub-neural network during the training process.
Still further, the processing based on the output signal and the first predicted output signal to obtain a corrected output signal operates to: calculating an absolute value diff of the difference between the output signal S and the first predicted output signal Ps, the corrected output signal Cs is:
Figure SMS_17
；
wherein ,
Figure SMS_18
the determination mode of (a) is as follows:
Figure SMS_19
as a result of the second threshold value being set,Tand 2 is a third threshold.
The invention also provides an output correction device of the optical fiber gyroscope based on the neural network under the low angular rate, which is characterized by comprising the following components:
the optical fiber gyroscope comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit acquires a current infrared image, working temperature and an output signal of the optical fiber gyroscope, and the output signal is generated by the optical fiber gyroscope at a low angular rate;
the prediction unit is used for inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, and the neural network is trained based on an optimized loss function;
a correction unit for processing the output signal and the first predicted output signal to obtain a corrected output signal;
the neural network comprises a first sub-neural network and a second sub-neural network, the first sub-neural network is trained based on sample infrared images in a training sample set, the second sub-neural network is trained based on sample working temperatures and sample output signals in the training sample set, and the first sub-neural network uses an loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function, wherein the optimized loss functionLThe method comprises the following steps:
Figure SMS_20
;
Figure SMS_21
;
Figure SMS_22
;
wherein ,
Figure SMS_23
representing the first of a training sample setiAn infrared image of the individual training samples is obtained,/>
Figure SMS_24
representing the first of a training sample setiWorking temperature of each training sample, +.>
Figure SMS_25
Representing the first of a training sample setiOutput signal of training sample, +.>
Figure SMS_26
Representing a supervision value->
Figure SMS_27
Respectively representPredictive value->
Figure SMS_28
Respectively represent the corresponding weight values.
Still further, filtering the acquired infrared image based on the operating temperature to discard unsuitable infrared images.
Still further, the filtering operation includes: and acquiring a standard infrared image corresponding to the working temperature from a database, calculating the similarity between the acquired infrared image and the standard infrared image, discarding the acquired infrared image if the similarity is smaller than a first threshold value, continuing to acquire the infrared image, and judging the acquired infrared image to be qualified if the similarity is larger than or equal to the first threshold value.
Further, determining
Figure SMS_29
The method comprises the following steps:
Figure SMS_30
；
Figure SMS_31
；
wherein ,
Figure SMS_32
representing the maximum value of the loss function of the first sub-neural network during training, < >>
Figure SMS_33
Representing the minimum value of the loss function of the first sub-neural network during training, +.>
Figure SMS_34
Representing the maximum value of the loss function of the second sub-neural network during training, < >>
Figure SMS_35
Representing the most significant loss function of the second sub-neural network during trainingSmall values.
Still further, the processing based on the output signal and the first predicted output signal to obtain a corrected output signal operates to: calculating an absolute value diff of the difference between the output signal S and the first predicted output signal Ps, the corrected output signal Cs is:
Figure SMS_36
；
wherein ,
Figure SMS_37
the determination mode of (a) is as follows: />
Figure SMS_38
As a result of the second threshold value being set,Tand 2 is a third threshold.
The invention also proposes an electronic device comprising a processor and a memory, the memory having stored thereon a computer program which, when executed by the processor, implements a method as claimed in any of the preceding claims.
The invention also proposes a computer readable storage medium having stored thereon computer program code which, when executed by a computer, performs any of the methods described above.
The invention has the technical effects that: the invention discloses an output correction method, a device and a storage medium under a low angular rate of a fiber-optic gyroscope based on a neural network, wherein the method comprises the following steps: the method comprises the following steps of S101, collecting a current infrared image, working temperature and an output signal of a fiber optic gyroscope, wherein the output signal is generated by the fiber optic gyroscope at a low angular rate; a prediction step S102, namely inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, wherein the neural network performs training based on an optimized loss function; a correction step S103 of processing the output signal and the first predicted output signal to obtain a corrected outputOutputting a signal; the invention provides a method for correcting the output of a fiber optic gyroscope at a low angular rate based on a neural network, which comprises a first sub-neural network and a second sub-neural network, wherein the first sub-neural network is trained based on a sample infrared image in a training sample set, the second sub-neural network is trained based on a sample working temperature and a sample output signal in the training sample set, and the first sub-neural network uses a loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function. That is to say, the neural network comprises two sub-neural networks, the two sub-neural networks can actually correct the output of the optical fiber gyroscope under the low angular rate, in order to improve the correction precision, after the two sub-neural networks are respectively trained independently, the two sub-neural networks are trained jointly based on the optimized loss function, so that the overall prediction performance of the two sub-neural networks is optimal, the specific optimized loss function is provided, the prediction performance is further improved, and the accuracy of the output correction of the optical fiber gyroscope under the low angular rate is improved; the invention provides an optimized loss function of two sub-neural network combined networks, wherein the key point of the loss function is a weight determination mode of two variables L1 and L2. The invention provides a specific correction formula, namely when the difference between the two is smaller than T1, the sum of the two is divided by 2, and when the difference between the two is larger than or equal to T1 and smaller than T2, the calculation is carried out based on the specific formula, and the calculation mode considers the effect of the absolute value diff of the difference between the two, so that the corrected signal is more accurate, and when the difference is larger than or equal to T2, the error of the predicted value is larger, the predicted value is required to be removed, and then the prediction is carried out again. In this way, the corrected output is improvedThe accuracy and reliability of the output signal.
Drawings
Other features, objects and advantages of the present application will become more apparent upon reading of the detailed description of non-limiting embodiments made with reference to the following drawings.
Fig. 1 is a flowchart of a method for correcting output at a low angular rate of a fiber optic gyroscope based on a neural network according to an embodiment of the present invention.
Fig. 2 is a block diagram of an output correction apparatus at a low angular rate of a fiber optic gyroscope based on a neural network according to an embodiment of the present invention.
Detailed Description
The present application is described in further detail below with reference to the drawings and examples. It is to be understood that the specific embodiments described herein are merely illustrative of the invention and are not limiting of the invention. It should be noted that, for convenience of description, only the portions related to the present invention are shown in the drawings.
It should be noted that, in the case of no conflict, the embodiments and features in the embodiments may be combined with each other. The present application will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
Fig. 1 shows a method for correcting output of a fiber optic gyroscope based on a neural network under a low angular rate, which is characterized in that the method comprises the following steps:
the method comprises the following steps of S101, collecting a current infrared image, working temperature and an output signal of a fiber optic gyroscope, wherein the output signal is generated by the fiber optic gyroscope at a low angular rate; the low angular rate refers to a small deflection angle of the optical fiber gyroscope per unit time, for example, less than 5 degrees/s, at the moment, the optical fiber gyroscope is considered to be at the low angular rate, and the output error generated at the low angular rate is larger because the output of the optical fiber gyroscope is greatly influenced by external conditions such as temperature, so that the output of the optical fiber gyroscope at the low angular rate is required to be corrected.
A prediction step S102, namely inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, wherein the neural network performs training based on an optimized loss function;
a correction step S103, processing based on the output signal and the first predicted output signal to obtain a corrected output signal;
the neural network comprises a first sub-neural network and a second sub-neural network, the first sub-neural network is trained based on sample infrared images in a training sample set, the second sub-neural network is trained based on sample working temperatures and sample output signals in the training sample set, and the first sub-neural network uses an loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function, wherein the optimized loss functionLThe method comprises the following steps:
Figure SMS_39
;
Figure SMS_40
;
Figure SMS_41
;
wherein ,
Figure SMS_42
representing the first of a training sample setiAn infrared image of the individual training samples is obtained,/>
Figure SMS_43
representing the first of a training sample setiWorking temperature of each training sample, +.>
Figure SMS_44
Representing the first of a training sample setiOutput signal of training sample, +.>
Figure SMS_45
Representing a supervision value->
Figure SMS_46
Respectively represent predicted values,/->
Figure SMS_47
Respectively represent the corresponding weight values.
The invention provides a method for correcting the output of a fiber optic gyroscope at a low angular rate based on a neural network, which comprises a first sub-neural network and a second sub-neural network, wherein the first sub-neural network is trained based on a sample infrared image in a training sample set, the second sub-neural network is trained based on a sample working temperature and a sample output signal in the training sample set, and the first sub-neural network uses a loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function. That is, the neural network of the application comprises two sub-neural networks, the two sub-neural networks can actually correct the output of the optical fiber gyroscope under the low angular velocity, and in order to improve the correction precision, the two sub-neural networks are respectively trained independently and then are trained jointly based on the optimized loss function, so that the overall prediction performance of the two sub-neural networks is optimal.
In a further embodiment, the acquired infrared image is subjected to a filtering operation based on the operating temperature to discard unsuitable infrared images. The filtering operation includes: and acquiring a standard infrared image corresponding to the working temperature from a database, calculating the similarity between the acquired infrared image and the standard infrared image, discarding the acquired infrared image if the similarity is smaller than a first threshold value, continuing to acquire the infrared image, and judging the acquired infrared image to be qualified if the similarity is larger than or equal to the first threshold value. The infrared image reflects the state of the optical fiber gyroscope at a certain temperature, so the optical fiber gyroscope is used as an input condition for calibration, however, the collected infrared image may be poor due to the interference of external factors, the collected image needs to be judged, the similarity between the collected infrared image and the standard infrared image is calculated, if the similarity is smaller than a first threshold value, the calculation can be performed through a specific calculation method of the image similarity, so the poor image is omitted, and the reliability of the calibration is improved.
In a further embodiment, a determination is made
Figure SMS_48
The method comprises the following steps:
Figure SMS_49
；
Figure SMS_50
；
wherein ,
Figure SMS_51
representing the maximum value of the loss function of the first sub-neural network during training, < >>
Figure SMS_52
Representing the minimum value of the loss function of the first sub-neural network during training, +.>
Figure SMS_53
Representing the maximum value of the loss function of the second sub-neural network during training, < >>
Figure SMS_54
Representing the minimum of the loss function of the second sub-neural network during the training process.
The invention provides a specific formula for determining weights based on the maximum value and the minimum value of the respective loss functions of two sub-neural networks in the training process, which is equivalent to setting the weights of the two sub-neural networks, improves the reliability of correction, and is another important invention point of the invention.
In a further embodiment, the processing based on the output signal and the first predicted output signal to obtain a corrected output signal is operative to: calculating an absolute value diff of the difference between the output signal S and the first predicted output signal Ps, the corrected output signal Cs is:
Figure SMS_55
；
wherein ,
Figure SMS_56
the determination mode of (a) is as follows:
Figure SMS_57
as a result of the second threshold value being set,Tand 2 is a third threshold.
In the prior art, a predicted value is generally directly used as a calibrated output signal, and the method is not suitable for output correction of the optical fiber gyroscope at a low angular rate, because the predicted value may be different from an actual output value greatly, and the reliability of the predicted value is low, in order to solve the problem, a specific correction formula is provided, namely, when the difference between the predicted value and the actual output value is smaller than T1, the sum of the predicted value and the actual output value is divided by 2, and when the difference between the predicted value and the actual output value is larger than or equal to T1 and smaller than T2, the calculation is performed based on the specific formula, and the effect of the absolute value diff of the difference between the predicted value and the actual output value is considered in the calculation mode, so that the corrected signal is more accurate, and when the difference between the predicted value and the predicted value is larger than or equal to T2, the error of the predicted value is required to be removed, and then the prediction is performed again. This way the accuracy and reliability of the corrected output signal is improved, which is an important inventive concept of the present invention.
Fig. 2 shows an output correction device of a fiber optic gyroscope based on a neural network under a low angular rate, which is characterized in that the device comprises:
the acquisition unit 201 acquires a current infrared image, a working temperature and an output signal of the optical fiber gyroscope, wherein the output signal is generated by the optical fiber gyroscope at a low angular rate; the low angular rate refers to a small deflection angle of the optical fiber gyroscope per unit time, for example, less than 5 degrees/s, at the moment, the optical fiber gyroscope is considered to be at the low angular rate, and the output error generated at the low angular rate is larger because the output of the optical fiber gyroscope is greatly influenced by external conditions such as temperature, so that the output of the optical fiber gyroscope at the low angular rate is required to be corrected.
The prediction unit 202 is used for inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, wherein the neural network is trained based on an optimized loss function;
a correction unit 203 that processes based on the output signal and the first predicted output signal to obtain a corrected output signal;
the neural network comprises a first sub-neural network and a second sub-neural network, the first sub-neural network is trained based on sample infrared images in a training sample set, the second sub-neural network is trained based on sample working temperatures and sample output signals in the training sample set, and the first sub-neural network uses an loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function, wherein the optimized loss functionLThe method comprises the following steps:
Figure SMS_58
;
Figure SMS_59
;
Figure SMS_60
;
wherein ,
Figure SMS_61
representing the first of a training sample setiAn infrared image of the individual training samples is obtained,/>
Figure SMS_62
representing the first of a training sample setiWorking temperature of each training sample, +.>
Figure SMS_63
Representing the first of a training sample setiOutput signal of training sample, +.>
Figure SMS_64
Representing supervision value, ->
Figure SMS_65
Respectively represent predicted values,/->
Figure SMS_66
Respectively represent the corresponding weight values.
The invention provides a method for correcting the output of a fiber optic gyroscope at a low angular rate based on a neural network, which comprises a first sub-neural network and a second sub-neural network, wherein the first sub-neural network is trained based on a sample infrared image in a training sample set, the second sub-neural network is trained based on a sample working temperature and a sample output signal in the training sample set, and the first sub-neural network uses a loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function. That is, the neural network of the present application includesThe invention aims to improve the correction precision, and the two sub-neural networks are respectively trained independently and then are trained based on an optimized loss function, so that the overall prediction performance of the two sub-neural networks is optimal.
In a further embodiment, the acquired infrared image is subjected to a filtering operation based on the operating temperature to discard unsuitable infrared images. The filtering operation includes: and acquiring a standard infrared image corresponding to the working temperature from a database, calculating the similarity between the acquired infrared image and the standard infrared image, discarding the acquired infrared image if the similarity is smaller than a first threshold value, continuing to acquire the infrared image, and judging the acquired infrared image to be qualified if the similarity is larger than or equal to the first threshold value. The infrared image reflects the state of the optical fiber gyroscope at a certain temperature, so the optical fiber gyroscope is used as an input condition for calibration, however, the collected infrared image may be poor due to the interference of external factors, the collected image needs to be judged, the similarity between the collected infrared image and the standard infrared image is calculated, if the similarity is smaller than a first threshold value, the calculation can be performed through a specific calculation method of the image similarity, so the poor image is omitted, and the reliability of the calibration is improved.
In a further embodiment, determine,
Figure SMS_67
The method comprises the following steps:
Figure SMS_68
;
Figure SMS_69
;
wherein ,
Figure SMS_70
representing the maximum value of the loss function of the first sub-neural network during training, < >>
Figure SMS_71
Representing the minimum value of the loss function of the first sub-neural network during training, +.>
Figure SMS_72
Representing the maximum value of the loss function of the second sub-neural network during training, < >>
Figure SMS_73
Representing the minimum of the loss function of the second sub-neural network during the training process.
The invention provides a specific formula for determining weights based on the maximum value and the minimum value of the respective loss functions of two sub-neural networks in the training process, which is equivalent to setting the weights of the two sub-neural networks, improves the reliability of correction, and is another important invention point of the invention.
In a further embodiment, the processing based on the output signal and the first predicted output signal to obtain a corrected output signal is operative to: calculating an absolute value diff of the difference between the output signal S and the first predicted output signal Ps, the corrected output signal Cs is:
Figure SMS_74
；
wherein ,
Figure SMS_75
the determination mode of (a) is as follows:
Figure SMS_76
as a result of the second threshold value being set,Tand 2 is a third threshold.
In the prior art, a predicted value is generally directly used as a calibrated output signal, and the method is not suitable for output correction of the optical fiber gyroscope at a low angular rate, because the predicted value may be different from an actual output value greatly, and the reliability of the predicted value is low, in order to solve the problem, a specific correction formula is provided, namely, when the difference between the predicted value and the actual output value is smaller than T1, the sum of the predicted value and the actual output value is divided by 2, and when the difference between the predicted value and the actual output value is larger than or equal to T1 and smaller than T2, the calculation is performed based on the specific formula, and the effect of the absolute value diff of the difference between the predicted value and the actual output value is considered in the calculation mode, so that the corrected signal is more accurate, and when the difference between the predicted value and the predicted value is larger than or equal to T2, the error of the predicted value is required to be removed, and then the prediction is performed again. This way the accuracy and reliability of the corrected output signal is improved, which is an important inventive concept of the present invention.
In one embodiment of the invention an electronic device is presented comprising a processor and a memory having a computer program stored thereon, which when executed by the processor implements the method of any of the above.
In one embodiment of the invention a computer storage medium is provided, on which a computer program is stored, which computer storage medium may be a hard disk, DVD, CD, flash memory or the like, which computer program, when being executed by a processor, carries out the above-mentioned method.
For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functions of each element may be implemented in one or more software and/or hardware elements when implemented in the present application.
Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.
Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel, sequentially or in a different order, provided that the desired results of the disclosed aspects are achieved, and are not limited herein.
Although embodiments or examples of the present disclosure have been described with reference to the accompanying drawings, it is to be understood that the foregoing methods, systems, and apparatus are merely exemplary embodiments or examples, and that the scope of the present invention is not limited by these embodiments or examples but only by the claims following the grant and their equivalents. Various elements of the embodiments or examples may be omitted or replaced with equivalent elements thereof. Furthermore, the steps may be performed in a different order than described in the present disclosure. Further, various elements of the embodiments or examples may be combined in various ways. It is important that as technology evolves, many of the elements described herein may be replaced by equivalent elements that appear after the disclosure.

Claims (10)

1. An output correction method of a fiber optic gyroscope based on a neural network under a low angular rate is characterized by comprising the following steps:
the method comprises the steps of collecting current infrared images, working temperature and output signals of the optical fiber gyroscope, wherein the output signals are generated by the optical fiber gyroscope at a low angular rate;
a prediction step, namely inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, wherein the neural network performs training based on an optimized loss function;
a correction step of processing based on the output signal and the first predicted output signal to obtain a corrected output signal;
wherein the method comprises the steps ofThe neural network comprises a first sub-neural network and a second sub-neural network, the first sub-neural network is trained based on sample infrared images in a training sample set, the second sub-neural network is trained based on sample working temperatures and sample output signals in the training sample set, and the first sub-neural network uses an loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function, wherein the optimized loss functionLThe method comprises the following steps:
Figure QLYQS_1
;
Figure QLYQS_2
;
Figure QLYQS_3
;
wherein ,
Figure QLYQS_4
representing the first of a training sample setiInfrared images of individual training samples, < >>
Figure QLYQS_5
Representing the first of a training sample setiWorking temperature of each training sample, +.>
Figure QLYQS_6
Representing the first of a training sample setiOutput signal of training sample, +.>
Figure QLYQS_7
The supervision value is indicated as such,
Figure QLYQS_8
separate tableShow predicted value +.>
Figure QLYQS_9
Respectively represent the corresponding weight values.
2. The method of claim 1, wherein the acquired infrared image is filtered based on the operating temperature to discard unsuitable infrared images.
3. The method of claim 2, wherein the filtering operation comprises: and acquiring a standard infrared image corresponding to the working temperature from a database, calculating the similarity between the acquired infrared image and the standard infrared image, discarding the acquired infrared image if the similarity is smaller than a first threshold value, continuing to acquire the infrared image, and judging the acquired infrared image to be qualified if the similarity is larger than or equal to the first threshold value.
4. The method of claim 3, wherein the step of,
determination of
Figure QLYQS_10
The method comprises the following steps:
Figure QLYQS_11
；
Figure QLYQS_12
;
wherein ,
Figure QLYQS_13
representing the maximum value of the loss function of the first sub-neural network during training, < >>
Figure QLYQS_14
Representing the first sub-neural network during trainingMinimum value of loss function ∈>
Figure QLYQS_15
Representing the maximum value of the loss function of the second sub-neural network during training, < >>
Figure QLYQS_16
Representing the minimum of the loss function of the second sub-neural network during the training process. />
5. An output correction device based on a neural network for a fiber optic gyroscope at a low angular rate, the device comprising:
the optical fiber gyroscope comprises an acquisition unit, a control unit and a control unit, wherein the acquisition unit acquires a current infrared image, working temperature and an output signal of the optical fiber gyroscope, and the output signal is generated by the optical fiber gyroscope at a low angular rate;
the prediction unit is used for inputting the acquired infrared image, the working temperature and the output signal into a trained neural network to perform prediction to obtain a first prediction output signal, and the neural network is trained based on an optimized loss function;
a correction unit for processing the output signal and the first predicted output signal to obtain a corrected output signal;
the neural network comprises a first sub-neural network and a second sub-neural network, the first sub-neural network is trained based on sample infrared images in a training sample set, the second sub-neural network is trained based on sample working temperatures and sample output signals in the training sample set, and the first sub-neural network uses an loss functionL1Training the second sub-neural network to use the loss functionL2Training, after the first and second sub-neural networks are trained, performing joint training on the first and second sub-neural networks by using the optimized loss function, wherein the optimized loss functionLThe method comprises the following steps:
Figure QLYQS_17
;
Figure QLYQS_18
;
Figure QLYQS_19
;
wherein ,
Figure QLYQS_20
representing the first of a training sample setiInfrared images of individual training samples, < >>
Figure QLYQS_21
Representing the first of a training sample setiWorking temperature of each training sample, +.>
Figure QLYQS_22
Representing the first of a training sample setiOutput signal of training sample, +.>
Figure QLYQS_23
The supervision value is indicated as such,
Figure QLYQS_24
respectively represent predicted values,/->
Figure QLYQS_25
Respectively represent the corresponding weight values.
6. The apparatus of claim 5, wherein the acquired infrared image is filtered based on the operating temperature to discard unsuitable infrared images.
7. The apparatus of claim 6, wherein the filtering operation comprises: and acquiring a standard infrared image corresponding to the working temperature from a database, calculating the similarity between the acquired infrared image and the standard infrared image, discarding the acquired infrared image if the similarity is smaller than a first threshold value, continuing to acquire the infrared image, and judging the acquired infrared image to be qualified if the similarity is larger than or equal to the first threshold value.
8. The apparatus of claim 7, wherein the device comprises a plurality of sensors,
determination of
Figure QLYQS_26
The method comprises the following steps:
Figure QLYQS_27
；
Figure QLYQS_28
；
wherein ,
Figure QLYQS_29
representing the maximum value of the loss function of the first sub-neural network during training, < >>
Figure QLYQS_30
Representing the minimum value of the loss function of the first sub-neural network during training, +.>
Figure QLYQS_31
Representing the maximum value of the loss function of the second sub-neural network during training, < >>
Figure QLYQS_32
Representing the minimum of the loss function of the second sub-neural network during the training process. />
9. An electronic device comprising a processor and a memory, the memory having stored thereon a computer program which, when executed by the processor, implements the method of any of claims 1-4.
10. A computer storage medium, characterized in that the computer storage medium has stored thereon a computer program which, when executed by a processor, implements the method according to any of claims 1-4.
CN202310380651.7A 2023-04-11 2023-04-11 Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network Active CN116086497B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310380651.7A CN116086497B (en) 2023-04-11 2023-04-11 Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310380651.7A CN116086497B (en) 2023-04-11 2023-04-11 Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network

Publications (2)

Publication Number Publication Date
CN116086497A true CN116086497A (en) 2023-05-09
CN116086497B CN116086497B (en) 2023-06-09

Family

ID=86206817

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310380651.7A Active CN116086497B (en) 2023-04-11 2023-04-11 Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network

Country Status (1)

Country Link
CN (1) CN116086497B (en)

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103593538A (en) * 2013-11-28 2014-02-19 东南大学 Fiber optic gyroscope temperature drift modeling method by optimizing dynamic recurrent neural network through genetic algorithm
CN110348563A (en) * 2019-05-30 2019-10-18 平安科技（深圳）有限公司 The semi-supervised training method of neural network, device, server and storage medium
CN112633267A (en) * 2020-12-11 2021-04-09 苏州浪潮智能科技有限公司 Method, system, device and medium for positioning text of picture
CN114754801A (en) * 2022-06-16 2022-07-15 北京理工导航控制科技股份有限公司 Method and device for compensating zero offset temperature of fiber-optic gyroscope based on neural network and storage medium
CN114973177A (en) * 2022-06-02 2022-08-30 广州小鹏自动驾驶科技有限公司 Parking space detection method and device, vehicle and storage medium
CN114973152A (en) * 2022-07-27 2022-08-30 北京华夏圣远能源科技有限公司 Method, device and medium for monitoring micromolecule recyclable fracturing fluid storage tank based on neural network
US20220309779A1 (en) * 2021-03-26 2022-09-29 Canon Kabushiki Kaisha Neural network training and application method, device and storage medium
CN115511069A (en) * 2021-06-22 2022-12-23 Oppo广东移动通信有限公司 Neural network training method, data processing method, device and storage medium

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103593538A (en) * 2013-11-28 2014-02-19 东南大学 Fiber optic gyroscope temperature drift modeling method by optimizing dynamic recurrent neural network through genetic algorithm
CN110348563A (en) * 2019-05-30 2019-10-18 平安科技（深圳）有限公司 The semi-supervised training method of neural network, device, server and storage medium
CN112633267A (en) * 2020-12-11 2021-04-09 苏州浪潮智能科技有限公司 Method, system, device and medium for positioning text of picture
US20220309779A1 (en) * 2021-03-26 2022-09-29 Canon Kabushiki Kaisha Neural network training and application method, device and storage medium
CN115511069A (en) * 2021-06-22 2022-12-23 Oppo广东移动通信有限公司 Neural network training method, data processing method, device and storage medium
CN114973177A (en) * 2022-06-02 2022-08-30 广州小鹏自动驾驶科技有限公司 Parking space detection method and device, vehicle and storage medium
CN114754801A (en) * 2022-06-16 2022-07-15 北京理工导航控制科技股份有限公司 Method and device for compensating zero offset temperature of fiber-optic gyroscope based on neural network and storage medium
CN114973152A (en) * 2022-07-27 2022-08-30 北京华夏圣远能源科技有限公司 Method, device and medium for monitoring micromolecule recyclable fracturing fluid storage tank based on neural network

Also Published As

Publication number Publication date
CN116086497B (en) 2023-06-09

Similar Documents

Publication Publication Date Title
CN115600891B (en) Big data analysis method and system applied to production monitoring of aqueous acrylic resin
CN116086497B (en) Output correction method, device and medium for optical fiber gyroscope under low angular rate based on neural network
CN112097772B (en) Robot and map construction method and device thereof
CN112113471A (en) Inertial navigation measurement trajectory correction method based on optimal fuzzy system
CN111626168A (en) Gesture recognition method, device, equipment and medium
CN116740145A (en) Multi-target tracking method, device, vehicle and storage medium
US6163291A (en) Successive approximation analog to digital converter with bit error correction
CN114614825B (en) Low-cost high-speed pulse signal data sampling and peak value detection method
CN116295528A (en) Nonlinear piecewise temperature compensation method for MEMS gyroscope
CN113962876A (en) Correction method and correction device for pixel distortion and terminal
CN112833912A (en) V-SLAM map checking method, device and equipment
US4206505A (en) Random process initial moments computer
CN112102942B (en) Skeletal development grade detection method and terminal equipment
CN115802355B (en) Mobile Internet of things card management method and device and cloud platform
JPH1075218A (en) Method and device for identifying characteristic boundary of data groups
CN117664117B (en) Drift data analysis and optimization compensation method for fiber optic gyroscope
JPH02503965A (en) Improved target association method
CN113395031B (en) Complex system fault diagnosis method and system based on multistage model
CN110580429B (en) Video background library management method, device and application thereof
CN113822270B (en) Method and device for determining abnormal value of high-frequency data of structural health monitoring
CN117406759B (en) Pipeline robot crawling calibration method and system
CN116908134B (en) Semi-quantitative analysis method for plasticizer content and training method for analysis model
JP3017504B2 (en) Expectation value extraction method for logic circuit simulation.
CN118067155A (en) Temperature compensation method and device for gyroscope
CN115540909A (en) Method for detecting and processing abnormal value of leveling current of inertial platform system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant