193 results about "Transmission matrix" patented technology

The invention discloses a cross-cultural customer service dialogue quality automatic evaluation method in combination with sentiment analysis, and relates to the technical field of natural language processing, and the method comprises the steps: carrying out the alignment of voice and text based on a transmission matrix in real time, extracting a speech, a metaphor and polarity, and generating a speech tag; constructing an emotion channel and a polite channel, and fusing expression and shielding intensity through sharing attention; comparing and aligning with the same language prototype in a regional culture baseline library to obtain a calibration representation and updating a language offset record table; the potential upgrading probability is represented and recurred according to round aggregation calibration, and a risk vector and a high-risk position are formed; fusing risk and business indexes by a capacity integral kernel, outputting a comprehensive quality score, and giving factors and round attributions; sample recovery is triggered according to score and feedback difference, a micro-weight training data set is constructed, gradient increment training is carried out under low-rank adaptation, and cross-language consistency, early recognition of upgrading risks and interpretable evaluation are achieved through a closed loop.

The invention relates to the technical field of electrical equipment detection, and discloses a terahertz time-domain spectroscopy-based GIS insulating material nondestructive testing method, which comprises the following steps: acquiring GIS insulating material time-domain signals by using a reflection-type terahertz time-domain spectroscopy (THz-TDS), and separating and reconstructing defect echoes based on electromagnetic simulation modeling in combination with wavelet transform, STFT and a transmission matrix method; multi-dimensional features are extracted, SVM and RBFNN are trained after dimensionality reduction, and a classification and quantitative regression model fusing physical constraints is constructed; and inputting a signal of a sample to be detected into the model to realize rapid, non-contact and accurate detection on the type, size and depth of the defect. The method overcomes the defects that traditional detection is low in efficiency, has damage or is difficult to quantify, the signal-to-noise ratio, the sensitivity and the automation level are remarkably improved, the method can be widely applied to GIS production quality control, installation acceptance, operation and maintenance monitoring, health assessment and life prediction, and powerful support is provided for safe and reliable operation of an electric power system.

The invention provides a multi-user near-field secure communication method based on stacked intelligent metasurfaces, and relates to the technical field of wireless communication. The method comprises the following steps: obtaining an equivalent beamforming matrix and a first-layer channel transmission matrix of an SIM according to parameters of the SIM in a base station; according to the transmission distance between the SIM and the user and the transmission distance between the SIM and the eavesdropper and the parameters of the SIM, modeling a near-field channel to obtain a near-field channel model; obtaining a data transmission rate in combination with a beamforming vector of the base station, and further determining the security and the rate; and by taking maximization of the safety and the rate as a target, constructing an optimization model and decomposing the optimization model into two sub-problems to carry out alternate iterative optimization so as to obtain an optimized equivalent beam forming matrix and an optimized beam forming vector. An SIM is introduced, the freedom degree of beam regulation and control is increased, optimization is carried out on the basis of a near-field channel model by taking maximization of safety and rate as targets, legal users and eavesdroppers are effectively distinguished, signal energy is accurately guided to the legal users, and the communication safety is improved.

The invention relates to the technical field of computer vision and multi-modal learning, in particular to a cross-modal label pool construction and classification method for weak supervision training tasks. The method comprises the following steps: constructing a labeled source domain data set {an image with a labeled frame, a frame label and a label prototype vector}; clustering the regional features of the target domain image to obtain a clustering center; optimizing the transmission matrix T according to the optimization target; giving a pseudo label to each cluster according to the transmission matrix T, and constructing a pseudo label pool {regional features of a target domain image, pseudo labels}; according to the method, distribution between the source domain category and the target domain cluster is introduced, probability distribution type alignment is achieved between the source domain category and the target domain feature cluster, false label category mismatching is effectively avoided, the covering capacity for low resource categories is improved, and the robustness of overall false supervision training is enhanced.

The invention provides a linear optical tweezer astigmatism field regulation and control method based on an optical transmission matrix. The design method comprises the following steps: establishing a 4 * 4 optical matrix for generating a linear optical tweezer astigmatism field and a transmission equation; analyzing the regulation and control rule of the parameters of the optical element on the astigmatism characteristics; collecting a fluorescence slice image of the initial astigmatic light field through an axial tomography scanning method; reconstructing the spatial distribution of the initial image light field; judging whether the light field accords with a preset astigmatism characteristic or not, if not, determining the astigmatism characteristic to be optimized, and adjusting element parameters according to a regulation and control rule; and acquiring the adjusted image of the astigmatism light field again, reconstructing the image, and continuously performing judgment iteration until the image accords with the preset astigmatism characteristic, thereby realizing regulation and control of the astigmatism light field. Matrix optics is used in optimization of a linear optical tweezer image dispersion light field optical system, the limitation of traditional trial and error design is avoided, the method can be applied to the fields of light control technology, light sheet fluorescence microscopic imaging and the like, and the performance of the optical system is improved.

The invention provides a power distribution information physical system reliability evaluation method and system, and the method comprises the steps: building a multi-dimensional information system network model which comprises an equipment operation state model, a packet loss model, a time delay model and an error code model, building a node adjacency matrix according to an equipment node connection relation, obtaining a node network transmission matrix based on the node adjacency matrix, and carrying out the reliability evaluation of a power distribution information physical system. Calculating expected values of availability of four attributes of the multi-dimensional information system; based on the calculated expected values of the availability of the four attributes of the multi-dimensional information system, a fault point area caused by manual operation due to unreliable information system performance and prolonged power failure time of the upstream and downstream of a fault point are obtained; generating a physical operation scene based on the prolonged power failure time and correcting the physical operation scene; and carrying out fault recovery decision calculation on the corrected physical operation scene to finally obtain a reliability evaluation index.

The invention discloses a chirp grating design method and device, equipment and a storage medium, and relates to the technical field of optical fibers, and the disclosed chirp grating design method comprises the steps: analyzing a preset target spectrum function to obtain a target grating structure parameter; layering the initial grating into a plurality of sub-grating regions with uniform periods based on the target grating structure parameters, and establishing a layered transmission matrix model; determining a refractive index modulation depth range of each sub-gate region according to the layered transmission matrix model; and performing iterative optimization on grating structure parameters of the initial grating according to the layered transmission matrix model, the target spectral function, the refractive index modulation depth range and a preset error function to obtain a target grating. The performance of the chirp grating can be optimized.

The embodiment of the invention provides a design method of a wave-absorbing / wave-transmitting electrically-controlled switchable antenna housing, the structure of the antenna housing comprises an absorbing layer frequency selective surface, a dielectric layer and a wave-transmitting layer frequency selective surface from top to bottom, the absorbing layer frequency selective surface can carry out in-band loss and in-band transmission, the wave-transmitting layer can carry out in-band transmission and out-of-band reflection, and the dielectric layer can carry out in-band transmission and out-of-band reflection. The overall structure interacts to form a radar radome structure integrating absorption and transmission; a microwave network is adopted to carry out impedance analysis on the absorption and transmission radome, parameters are set to represent the characteristics of the microwave network, and meanwhile, a transmission matrix of the microwave network is analyzed through a matrix. The designed antenna housing not only has good out-of-band wave-absorbing stealth performance, but also can switch an in-band wave-transmitting state to a wave-absorbing state through electric control to realize limit stealth.

The invention provides a low-altitude unmanned aerial vehicle visual target tracking method based on dynamic sparse Sinkhorn attention, and belongs to the technical field of unmanned aerial vehicle intelligent perception. According to the method, an unmanned aerial vehicle carries imaging equipment to obtain video images of a target template area and a search area, and visual and language tokens are respectively generated in combination with natural language description; a dynamic sparse Sinkhorn attention module is embedded behind a multi-head self-attention layer of a visual encoder, an approximate double-random transmission matrix is constructed through similarity calculation and iteration, and differential sorting between tokens is achieved; a previous visual token most related to target semantics is dynamically screened and reserved according to the correlation score, background interference and redundant information are inhibited, and airborne calculation burden is reduced; and fusing the screened tokens in a prediction network, and outputting a target existence probability and a bounding box to realize stable real-time tracking. According to the method, the calculation complexity is reduced on the premise of ensuring the precision, and a low-power-consumption embedded platform is adapted; the robustness in a complex low-altitude environment is enhanced; and the cross-domain generalization ability is improved.

The invention provides a high-voltage circuit breaker fault diagnosis method based on voiceprint noise reduction and cloud edge cooperation. The method comprises the following steps: firstly, acquiring environmental background noise in an idle period of the high-voltage circuit breaker, and synchronously acquiring multichannel original voiceprint signals in an operation process; and through covariance modeling and relative transmission matrix calculation, noise reduction processing of multi-channel voiceprint signals is completed, and pure voiceprint signals are extracted. Then, performing feature extraction on the pure voiceprints, screening voiceprint channels which are most relevant to the fault state and are high in complementarity based on a mutual information criterion, and constructing a representative voiceprint set; and uploading the representative voiceprint set to a cloud end, and realizing fault type identification and positioning through a trained intelligent model. The method combines the advantages of edge real-time processing and cloud deep analysis, can effectively improve the accuracy and response speed of fault diagnosis of the high-voltage circuit breaker, and has high reliability and practical value.

The invention discloses a partial discharge harmonic online early warning method for a cable branch box, and particularly relates to the technical field of electrical measurement and power equipment state monitoring. And a unified time base is difficult to obtain through multi-point monitoring, harmonic distortion is easy to be misjudged as discharge, and stable positioning is difficult in a topological constraint. Harmonic scene fingerprints are constructed under a unified time base, version locking is carried out, power frequency phases are re-calibrated according to the fingerprints, gating compensation is set in a non-whole frequency band domain, and joint matching is carried out by combining arrival time difference with energy attenuation under the constraint of a topological equivalent transmission matrix. And dual-threshold interlocking trigger early warning is carried out by combining the separation degree with topological consistency, so that partial discharge pulses can still be stably displayed under the change of a harmonic working condition, false alarm and caliber drift caused by harmonic waves are inhibited, reviewable positioning and executable graded early warning of branch and joint levels are realized, and the reliability of the system is improved. And the stability, comparability and implementability of field judgment are improved.

The invention discloses a heterogeneous point cloud registration method. The method comprises the following steps: S1, preprocessing; s2, extracting high-scale aggregation features; s3, inputting the features into a network, and through calculation of self-attention and cross attention, combining point-to-angle feature embedding vectors to obtain enhanced features; s4, taking the enhanced features as input, calculating a feature similarity matrix, and performing iterative calculation through an optimal transmission algorithm to obtain an optimal transmission matrix; s5, calculating a global scale factor by taking the optimal transmission matrix as input, and executing scale alignment operation; and S6, taking the optimal transmission matrix and the two groups of scale alignment point clouds as input, and carrying out point cloud accurate registration. According to the method, the point cloud registration error is reduced, the difference between the point features is increased, the parameter solving complexity is reduced, and the overall operation efficiency is remarkably improved on the premise of ensuring the precision.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as financial science and technology and medical health, and discloses a task processing method, device and equipment based on optimal transmission and a medium. Extracting visual, audio and motion features by using a plurality of modal encoders, and generating feature representations of a plurality of modals; constructing an inter-modal feature distribution matrix based on the feature representation; generating a transmission matrix through an input convex neural network and an entropy regularization iteration mode; realizing modal feature distribution alignment by using the transmission matrix to obtain aligned feature representation; adopting a gated cross attention mechanism to fuse and generate unified feature representation; and executing the target task based on the unified feature representation, and outputting a task processing result. According to the method, modal feature alignment is realized by inputting the convex neural network in combination with entropy regularization, and fusion is completed by using gated cross attention, so that cross-modal semantic consistency and calculation efficiency are improved.

The invention discloses a recommendation model training method and device based on article fairness. The recommendation model training method comprises the steps that ID embedded representations of a user and an article and multi-modal content embedded representations of the article are constructed; a cost matrix is constructed based on the multi-modal semantic similarity between the articles, source distribution and target distribution in optimal transmission are redefined by using a weighting mechanism and a fairness transmission mechanism based on article exposure, and optimal transmission loss is constructed based on the cost matrix and a transmission matrix; and constructing ID representation loss based on the ID embedded representation of the user and the article and the interaction label, constructing multi-modal content representation loss based on the ID embedded representation of the user and the multi-modal content embedded representation of the article and the interaction label, and constructing the three types of loss into joint loss to train an article recommendation model. The method can effectively improve feature representation learning of unpopular articles, realizes collaborative improvement of fairness and performance of a recommendation system, and is especially suitable for cold start and long-tail article recommendation scenes.

The invention provides a multi-camera image processing method and system, electronic equipment and a medium, and belongs to the technical field of aerial view, and the method comprises the steps: obtaining fisheye images shot by a plurality of fisheye cameras; converting the image features of each fisheye image into corresponding aerial view features through a feature transmission matrix; and performing aerial view feature fusion through a distortion perception learning pooling layer. According to the method, the advantage of a large field angle of the fisheye camera is utilized, efficient feature conversion is achieved in combination with matrix multiplication, feature fusion is optimized through a distortion perception pooling layer, and the problem of radial image distortion of the fisheye camera is effectively solved; according to the invention, through feature compression and matrix decomposition technologies, the calculation amount and memory occupation of multi-camera BEV feature fusion are greatly reduced; the shielding reasoning module is introduced, and the shielding problem in the BEV space can be effectively solved.

The invention relates to the technical field of high-speed aircraft ground simulation and electromagnetic property calculation, and provides a high-speed aircraft electromagnetic calculation method and system based on a quantum classification algorithm. A variable component sub-classifier is used as an intelligent scheduling core to perform feature recognition and region division on a whole flow field, and a transmission matrix method and a WKB approximation method are adaptively called according to a division result to perform hybrid calculation, so that the technical problem that a traditional single calculation strategy is difficult to collaboratively optimize in precision and efficiency is solved; the method has the advantages that global computing resources can be distributed according to needs, and the overall solving speed is greatly increased while high simulation precision of the key area is guaranteed.

The present application relates to the field of data processing technology, and in particular to a multi-source data fusion method and system based on an optimal transmission algorithm, the method comprising obtaining an input data source and extracting a feature vector of each data source; merging the feature vectors of each data source and performing a clustering operation to generate a prototype set, defining a joint weight distribution of multi-path jump paths by calculating the jump probability between a cross-modal vector and a vector in the prototype set, obtaining an entropy-weighted multi-path prototype bridging cost that fuses multiple semantic path costs based on the joint weight distribution; constructing a sparse optimal transmission matrix based on the optimal transmission algorithm and the entropy-weighted multi-path prototype bridging cost; and utilizing the sparse optimal transmission matrix to perform cross-modal weighted alignment and enhanced representation construction on the extracted feature vectors of each data source to generate a multi-modal fusion representation. The present application can reduce the consumption of computing resources while ensuring the accuracy and efficiency of data fusion.

The invention discloses a welding spot testing method and system and a computer program product, and relates to the technical field of electronic packaging, and the method comprises the steps: constructing a single welding spot testing architecture through a to-be-tested single welding spot of a chip unit, and determining a transmission parameter matrix according to a scattering parameter matrix of the single welding spot testing architecture; combined de-embedding calculation is carried out by combining a substrate open circuit transmission matrix, a substrate short circuit transmission matrix and a chip short circuit transmission matrix which are measured when the single welding spot test framework is in an open circuit structure and a short circuit structure, so that the parasitic effect of the test substrate and the chip unit and the crosstalk of adjacent structures are accurately stripped; therefore, transmission performance parameters which only reflect intrinsic characteristics of the to-be-tested single welding spot can be directly extracted, signal superposition interference and approximate estimation limitation existing in a traditional multi-point series connection test or parameter equalization method are broken through, systematic errors are eliminated, accurate measurement and fault location of electrical characteristics of the to-be-tested single welding spot are achieved, and the method is suitable for large-scale popularization and application. And the accuracy and the reliability of the welding spot test in the packaging body are obviously improved.

A computing apparatus is provided. The apparatus includes: an input port to receive a first optical signal; a matrix calculation unit having a transmission matrix and configured to use the transmission matrix to process the first optical signal to obtain a second optical signal, where the second optical signal has a complex amplitude, and the transmission matrix is a unitary matrix M′; a real part obtaining unit optically coupled to the matrix calculation unit and configured to obtain, from the second optical signal, a third optical signal representing a real part of a complex amplitude of the second optical signal; and an output port configured to output an optical signal.

The invention belongs to the technical field of electromagnetic field calculation, and discloses a method and a system for calculating an electromagnetic scattering field of a moving target coated with a lossy dispersive medium. The method comprises the following steps: firstly, carrying out three-dimensional modeling on a target surface by adopting triangular surface elements to form a model file, and importing the model file into a main program to obtain geometric feature data of a target; secondly, carrying out relativistic coordinate transformation on the space-time component by utilizing Lorentz transformation, and mapping a dynamic problem to an inertial coordinate system synchronous with a target; then establishing an equivalent reflectivity model for the lossy dispersive medium coated area by adopting a transmission matrix method; then efficiently solving a transient scattering field in the motion coordinate system by adopting a time domain physical optical method; and finally, reconstructing a calculation result to a laboratory coordinate system through inverse Lorentz transformation, and outputting electromagnetic field distribution in a dynamic scene. Through the method provided by the invention, the electromagnetic scattering field of the moving target coated with the lossy dispersive medium can be quickly calculated.

This invention relates to a method, apparatus, device, and medium for error compensation in a dual-drive feed system, belonging to the field of CNC equipment error measurement and compensation technology. The method includes: establishing a low-order volume array of the equipment to be modeled, equipped with a dual-drive feed system; obtaining an error transmission chain from the workpiece to the tool based on the low-order volume array; establishing a mathematical model of tool error and dual-axis error when the dual axes of the dual-drive feed system drive the tool; determining the homogeneous coordinate transformation matrix between any two volumes in the low-order volume array; establishing an error transmission matrix from the workpiece coordinate system to the tool coordinate system based on the error transmission chain, mathematical model, and homogeneous coordinate transformation matrix; establishing a comprehensive error model of the dual-drive feed system based on the error transmission matrix and the transmission matrix under ideal conditions; and performing error compensation on the dual-drive feed system based on the comprehensive error model, thereby improving the error compensation accuracy of the dual-drive feed system.

The application discloses a wave plate cascade disturbance device and method based on outer rotation and inner rotation forms, and the device comprises an optical transmitter, a disturbance module and a calculation module. The method comprises the following steps: acquiring polarized light with a fixed polarization state; determining a transmission matrix of an outer rotation model and a transmission matrix of an inner rotation model given a disturbance speed and a white noise distribution; performing decomposition conversion processing on the transmission matrix of the outer rotation model and the transmission matrix of the inner rotation model to obtain three-level electrical signals; and performing polarization state adjustment processing on the polarized light with the fixed polarization state according to the three-level electrical signals to obtain adjusted polarized light. The application can independently control the disturbance speed and the polarization probability distribution, improve the controllability of the disturbance model, and realize high-speed polarization disturbance with strong adjustability. The application can be widely applied to the technical field of polarized light disturbance.

The application discloses a wide-angle optical antireflection method based on equivalent gradient refractive index matching and phase angle restriction, and the method can solve the problem that some medium materials corresponding to the refractive index do not exist, so as to meet the needs of subsequent gradient refractive index matching, by selecting two medium materials with different refractive indexes and respectively controlling the thickness of the medium layers, and converting the medium layers into equivalent medium layers with different equivalent refractive indexes and equivalent thicknesses according to a transmission matrix. In addition, the change of the propagation phase is restricted by limiting the propagation angle of the incident light in the medium layer, so as to realize the insensitivity of the propagation phase to the angle change, meet the needs of wide-angle transmission enhancement, and determine the medium material corresponding to the refractive index according to the range of the propagation angle restriction. According to the above method, the medium material capable of realizing the wide-angle optical antireflection can be quickly selected, and the time cost required for selecting the medium material is greatly shortened.

The application relates to a double-symmetry P-B phase high-transmittance super surface structure sequence and a unit pattern design method, and belongs to the field of artificial electromagnetic super materials. The designed double-symmetry P-B phase high-transmittance super surface is a coded super surface which is sequentially stacked by a pattern layer, a transparent dielectric layer and a metal mesh layer. The application proposes a method for realizing the design of the geometric parameters of the pattern layer meshed unit pattern patch based on a transmission matrix and an equivalent circuit method, greatly reduces the design time, simplifies the design process, and can be applied to the design of pattern patches in different target frequency bands. Meanwhile, the super surface structure sequence is obtained by optimizing the fitness function about the multi-point scattering characteristics in the target frequency band, so that the coded super surface has low scattering performance in the target frequency band. In addition, the designed super surface can realize wide-band high-transmittance, is widely applied to high-performance light windows of aerospace, medical treatment and precision instrument equipment, and has a wide application prospect.

The application discloses a mechanical equipment safety interlock control method based on multi-sensor information fusion. The method collects the motion position, speed, acceleration, force and torque data of the mechanical equipment, and performs time alignment, amplitude normalization and short-time abnormality filtering, estimates the kinetic energy, potential energy and friction energy consumption of each energy node, and constructs an energy transmission matrix between nodes; fusion sensor data and matrix information extracts four types of abnormal characteristics of short-term energy surge, local energy reflection, friction energy abnormal fluctuation and energy accumulation accompanied by micro-vibration, and predicts dangerous energy states including single and complex dangerous energy states accordingly; based on the prediction of dangerous energy state and the current running state of the equipment, the safety interlock action is adaptively triggered, the driver torque, current amplitude limiting and active damping are dynamically adjusted, and the control of backflow energy and accumulated energy along the virtual dissipation path is realized. The method is suitable for high dynamic mechanical equipment and can realize fine safety interlock control of multiple nodes and multiple abnormal characteristics.

The invention provides an air interface wire rapid calibration method and system, and solves the problems of high execution complexity and difficulty in engineering application caused by the adoption of space violent search in the existing air interface wire calibration technology. The method comprises the following steps: synchronously starting a static calibration antenna and a reference antenna fixedly connected with a rotary table under a plurality of rotation angles of the rotary table, and acquiring measurement data from measured equipment; determining a phase difference data set and a relative amplitude data set according to the measurement data; reconstructing a plurality of air interface transmission matrixes in a digital domain based on the data set, and screening out at least one calibration antenna spatial position combination by taking a matrix condition number as an index; and air interface wire calibration is completed according to the combination. According to the invention, the calibration complexity is reduced from a combination order to a linear order, the hardware cost and the measurement time are significantly reduced, the high-efficiency and high-precision calibration of large-scale terminals such as a whole vehicle is realized, and the method has wide applicability irrelevant to frequency.

The invention provides a parametric modeling method and device for a silicon-based millimeter wave on-chip component, and belongs to the technical field of integrated circuit modeling and millimeter wave circuit design. The method comprises the following steps: under a preset key parameter interpolation node, acquiring a transmission parameter of a de-embedding test structural member comprising a de-embedded component at each test frequency point; removing a parasitic effect embedded in the test structural member by using transmission matrix operation so as to obtain a transmission matrix of the de-embedded component and obtain a scattering parameter of the de-embedded component; then, through a vector fitting algorithm, obtaining a pole-residue model of the de-embedded component under the parameter interpolation node; and based on the pole-residue model under each parameter interpolation node, establishing a parameterized de-embedding model of the de-embedded component through a gravity center interpolation algorithm. According to the method, rapid and accurate modeling of silicon-based millimeter wave on-chip components of any size can be realized, and the reliability and efficiency of millimeter wave integrated circuit design are improved.

The application discloses a kind of multi-mode optical fiber wavelength multiplexing wide field imaging quality promotion methods based on dynamic calibration.The method comprises the following steps: collecting the output speckle pattern corresponding to each wavelength after being transmitted by multi-mode optical fiber;The speckle patterns of different wavelengths are spatially registered and spliced to form a high-dimensional composite speckle vector;PCA is used to bidirectionally compress the dimension of input image and composite speckle vector group, and a linear wavelength multiplexing inverse transmission matrix (WITM) is dynamically estimated and updated in low-dimensional space, and real-time pre-reconstruction and continuous monitoring are realized;Set up optical fiber state monitoring module to analyze pre-reconstruction frame in real time, when structural disturbance or optical fiber deformation is detected, automatically trigger in situ dynamic calibration, realize the rapid adaptive update of WITM, without retraining network.The updated WITM will guide the diffusion neural network image optimizer to iteratively denoise and detail repair the pre-reconstruction result, significantly improve the image fidelity and structure restoration ability.

The invention provides a programmable photon integrated circuit phase configuration optimization method and system based on a neural network, and relates to the technical field of photon signal processing. The method comprises the following steps: firstly, randomly generating multiple groups of phase configurations in a chip, measuring to obtain a corresponding transmission matrix, and constructing a sample data set of phase configuration-matrix elements; training a full-connection neural network based on the data set to enable the full-connection neural network to accurately fit transmission characteristics under a component error condition; taking the phase configuration to be optimized as a parameter to be optimized, iteratively calculating loss through a gradient descent algorithm, and updating the parameter; and when the loss meets a preset condition, outputting the target phase configuration and loading the target phase configuration to the chip, thereby realizing high-precision, low-cost and directly applicable phase configuration optimization of the MZI network.

The application relates to a photoetching calibration device, a photoetching calibration method and a photoetching system for the photoetching system, wherein the photoetching calibration method comprises the following steps: a calibration objective lens collimates a first scattered light beam into a calibration collimated light beam; wherein a projection objective lens converges a to-be-modulated light beam emitted by a light source on a scattering medium, the to-be-modulated light beam is scattered into the first scattered light beam through the scattering medium; the distance between the calibration objective lens and the scattering medium is a preset focusing distance; a calibration converging lens converges the received calibration collimated light beam into a calibration light beam; a calibration camera receives the calibration light beam and takes a picture to obtain a calibration image; a calibration controller generates a scattering medium transmission matrix according to the calibration image, and calibrates the photoetching system based on the scattering medium transmission matrix. Through the application, the problem of how to reduce the production cost while ensuring that the photoetching pattern meets the high-precision requirement is solved, and the photoetching effect with low cost and high resolution is realized.