244 results about "Dynamic modulation" patented technology

The invention belongs to the technical field of cleaning and dirt prevention and removal, and discloses a control system and a control method of an ultrasonic cleaning device, stable energy coupling and dead-corner-free cleaning are realized through double-closed-loop driving and dynamic modulation in an ultrasonic cleaning link, and the uniformity of tank liquid cavitation intensity is remarkably improved. Meanwhile, control strategies are customized in the aspects of a rinsing link, a water cutting and drying link, liquid level control, manipulator control and the like, so that the defects of high energy consumption, incomplete drying, low rinsing efficiency, manipulator conflict and the like in the prior art are overcome while the surface and the interior of the workpiece are thoroughly cleaned; and the cleaning effects of high cleanliness, complete drying and no physical or chemical damage are ensured.

The invention provides a geometric parameter measurement method based on underwater structure crack pixel-level semantic segmentation, and the method comprises the steps: constructing a multi-level feature extraction network based on ResNet-34, providing a structured feature pyramid for the subsequent context modeling and interference suppression, modeling illumination distribution features through a multi-head self-attention mechanism, and carrying out the modeling of the illumination distribution features through a multi-head self-attention mechanism. Regional scattering noise is extracted in combination with a cavity convolution pyramid, an environment context tensor is generated through fusion, a learnable gating mechanism is introduced for a decoding stage, dynamic modulation of jump connection characteristics and interference area response suppression are achieved, in addition, an edge perception auxiliary loss function is adopted, topological continuity of crack edges is enhanced, and finally, the method is applied to the field of jump connection. High-precision pixel-level segmentation of the underwater crack area can be realized without prior illumination correction, and the detection robustness and segmentation integrity in a complex environment are improved.

The invention relates to the technical field of data processing and mode recognition, in particular to a bearing fault diagnosis method based on multi-scale feature fusion, which comprises the following steps: fusing multi-source data such as vibration, acoustic emission and rotating speed, performing angle domain resampling by using rotating speed data, generating a two-dimensional order spectrogram, and stacking to construct a three-dimensional working condition information tensor; a master-slave modulation heterogeneous neural network is adopted, high-dimensional spatial-temporal features are extracted through a main branch three-dimensional convolutional network, time sequence details are extracted from an original sequence through an auxiliary branch one-dimensional convolutional network, affine transformation parameters are generated, and dynamic modulation is achieved on the high-dimensional features; and the output state vector is mapped to a fault evolution knowledge graph, probability prediction is carried out through a graph attention network and by introducing a Monte Carlo discarding mechanism, a probability mean value is calculated as a fault classification result, and the diagnosis confidence is quantified by a probability variance. According to the invention, through multi-scale feature fusion and dynamic modulation, the problem of insufficient feature discrimination caused by scale mismatch under variable working conditions is solved.

The invention discloses an intelligent data mining system and method based on big data service, and aims to effectively fuse instant interests and historical preferences of users by introducing a short-term intention feature vector and a long-term user preference portrait coding vector and utilizing a fine-grained semantic comparison network to realize deep interaction between the short-term intention feature vector and the long-term user preference portrait coding vector. On the basis, the recall quantity proportion of the first recall channel and the second recall channel is designed to be dynamically modulated according to the comparison result between the short-term intention feature vector and the long-term preference portrait coding vector. In this way, the recall weights of different types of commodities or contents can be adaptively adjusted according to the consistency or difference of the current demands and historical interests of the users in different scenes. The dynamic modulation mechanism significantly improves the adaptability of the recommendation system to diversification and real-time performance in a complex scene, and is an important technical breakthrough for future intelligent services.

The invention relates to the technical field of beacon light control, in particular to a beacon light anti-interference control method and system based on environmental perception and dynamic modulation. The method comprises the following steps: extracting an ambient light reference based on adaptive weight sliding filtering; in combination with target detection and trajectory prediction, interference pre-judgment features are obtained; on the basis of an evidence theory, analyzing comprehensive interference evaluation characteristics through the ambient light reference, the interference pre-judgment characteristics and the electromagnetic interference characteristics; fusing meteorological environment perception and influence features and self health state features, and evaluating equipment operation risks; and based on modulation signal optimization characteristics, the comprehensive interference evaluation characteristics and the equipment operation risk, fuzzy control and rule reasoning are utilized to realize dynamic output of anti-interference control parameters. In order to solve the problem that the signal recognition rate is reduced due to the fact that a beacon light is easily affected by multiple factors in a complex ocean navigation channel environment, data are obtained through multi-dimensional environment perception, and dynamic modulation control is achieved after feature extraction and fusion.

The invention discloses an ultra-high-speed imaging device based on acousto-optic filtering modulation. The ultra-high-speed imaging device comprises a spectrum modulation system, a spectrum imaging system and a data processing system. The spectrum modulation system utilizes an acousto-optic tunable filter (AOTF) to dynamically modulate broadband continuous spectrum light, narrow-band illumination light which rapidly changes along with time is generated, and time information of a dynamic scene is effectively mapped to a spectrum dimension. And the spectral imaging system images the spectral coded dynamic scene to the hyperspectral camera to realize acquisition of spectral space-time information. The data processing system performs spectral crosstalk correction, channel separation and demosaicing reconstruction on the acquired original mosaic image, and recovers a time sequence image according to a spectrum-time mapping relation, thereby realizing single-shot superspeed imaging in a dynamic process. The device has the advantages of being simple and compact in structure, flexible in time window, high in imaging fidelity and the like, and has wide application prospects in the fields of rapid dynamics such as microfluidics, laser processing and ultrafast physics.

The invention discloses a dynamic digital twinning three-dimensional Gaussian splash rendering system and a dynamic digital twinning three-dimensional Gaussian splash rendering method. According to the system, a static scene is reconstructed by utilizing a three-dimensional Gaussian splashing technology through a semantic scene reconstruction module, semantic information is associated to Gaussian primitives in combination with spatial alignment and semantic injection, and a model with a semantic identifier is generated; multi-source dynamic data are fused and converted into a dynamic data field which can be accessed by a GPU through a multi-mode spatio-temporal data processing module; and through a rendering attribute dynamic modulation module, screening a target primitive according to the semantic identifier, and directly modulating internal rendering attributes such as color, transparency or shape of the target primitive based on a data field query value, thereby realizing internal rendering and visualization of dynamic data. According to the method, the problems that in the prior art, rendering reality and real-time performance are difficult to give consideration to and data and scene fusion is superficial are solved, and dynamic digital twinning rendering with high reality and strong immersion is realized.

The invention discloses a driver state remote monitoring collaborative alarm platform and method based on the Internet of Things, and relates to the technical field of the Internet of Things, and the method comprises the steps: carrying out the channel information entropy analysis of collected state data, adaptively cutting off a low-information-amount channel, carrying out the nonlinear feature mapping to obtain a mapping feature sequence, constructing a space-time coupling information tensor, and carrying out the self-adaptive cutting of a low-information-amount channel; the method comprises the following steps: calculating a multi-mode cooperative surge factor, executing tensor decomposition to obtain a decomposition factor, reconstructing an estimation tensor and calculating a reconstruction error, calculating a multi-mode entropy surge factor, generating an abnormal score, counting and generating a dynamic threshold according to the abnormal score in a sliding window, and judging a real-time score to generate an alarm trigger signal. Through adaptive dimensionality reduction, hybrid mapping, dynamic modulation CP decomposition and adaptive reconstruction error evaluation based on channel information entropy, real-time efficient processing and deep multi-modal fusion of multi-source data are realized, noise and redundancy are effectively eliminated, abnormal features are amplified, space-time coordination is accurately described, and dynamic adaptive anomaly judgment is realized.

The invention relates to the field of metal plastic forming control, and discloses a self-adaptive control system and method for forging and pressing aviation precision forgings, and the system comprises an online dynamic model identification module which identifies a process fingerprint representing the characteristics of a current blank at the initial stage of machining; the target curve self-adaptive regulator is used for regulating a reference energy dissipation rate curve according to the fingerprint and generating a personalized reference curve; the model prediction controller performs rolling optimization by using continuously updated model parameters and taking the reference curve as a tracking target, and outputs an optimal speed instruction; and the event triggering and constraint dynamic modulation module is used for actively tightening the operation constraint applied to the controller before a forging pressure sudden increase event is pre-judged through a parallel monitoring model prediction error. According to the method, through combination of target self-adaption and constraint dynamic modulation, control over the forging and pressing process is achieved, forging pressure overshoot can be effectively restrained, and the forming precision and batch-to-batch quality consistency of forge pieces are remarkably improved.

The invention provides a double-optical-comb-cavity modulated gas measurement system and method based on electro-optical modulation, in the use process, a detection optical comb and a reference optical comb which are coherent are generated through a double-optical-comb output module, the detection optical comb enters a resonant cavity to interact with gas to be measured, and the reference optical comb directly reaches a data acquisition and analysis module; the data acquisition and analysis module is communicated with the resonant cavity and the double-optical-comb output module, and is used for receiving the reference optical comb and the detection optical comb output from the resonant cavity, and performing component and concentration analysis on the to-be-detected gas according to the detection optical comb and the reference optical comb; in the measuring process, the electro-optical modulation module is used for acquiring the detection optical comb output from the resonant cavity in real time, and the optical field of the detection optical comb in the resonant cavity is dynamically modulated based on the detection optical comb, so that the optical comb mode is matched with the free spectral range of the resonant cavity in real time, and efficient coupling is formed in the resonant cavity; the influence of the gas dispersion effect on the measurement precision is eliminated.

The invention provides a video content enhancement method for a low-illumination environment, and the method comprises the steps: achieving the data preprocessing based on an original low-illumination video frame sequence through frame synchronization, color space conversion and local brightness analysis, generating a noise sensitivity thermodynamic diagram through multi-feature unsupervised learning, and constructing a noise perception gating mechanism through the combination of affine transformation. Dynamic modulation of the characteristic channel is realized; in the multi-scale network structure, a channel attention module is used for carrying out layer-by-layer self-adaptive adjustment on a noise sensitive area; a basic illumination image and an edge enhancement image are generated through double-branch decoding, and then weighted fusion is carried out in combination with a noise thermodynamic diagram, so that brightness balance and detail enhancement are realized; a noise smoothing regular term is introduced during end-to-end training, so that the network achieves dynamic balance between an enhancement effect and noise control.

The invention discloses a display light color detection system, and relates to the technical field of low-brightness display light color detection, and the system comprises a dynamic modulation module which is configured to carry out the high-frequency pulse modulation of a dark-state signal of a detection object, and is used for enabling the dark-state signal to carry a modulation component of a characteristic frequency; the time sequence division acquisition unit comprises a high-speed optical shutter assembly and a detector, and the high-speed optical shutter assembly synchronously switches the on-off state of an optical path according to the characteristic frequency; according to the invention, through high-frequency pulse modulation of dark state signals, time-sequence-divided acquisition of noise and signals and time-frequency domain combined decoupling, traditional hardware limitation is broken through, ultra-low brightness signals and stray light interference are effectively separated, 0.01 cd / m < 2 >-level high-precision detection is realized, the detection efficiency is improved by more than 7 times, and the misjudgment rate is significantly reduced. The problems of missing detection of a novel display technology and the yield of an automatic production line are solved, and an active coding-time sequence calibration-cross-domain cooperation framework has remarkable innovativeness and industrial value.

The invention discloses a geometric error identification and compensation system and method for a numerical control machine tool, and belongs to the technical field of numerical control machine tools, and the method comprises the steps: collecting the position, temperature and cutting force data of the machine tool in real time, and calculating the components of a position error, a thermally induced equivalent error and a force induced equivalent error to construct a multi-dimensional state vector; according to the method, a compensation value is calculated through a dynamic compensation model, the core of the method is that a dynamic modulation factor is introduced, the factor integrates the heat effect, the force effect and the multi-axis coupling effect to dynamically adjust the compensation weight, in addition, a grading trigger compensation strategy based on comprehensive error amplitude evaluation is designed, and the compensation accuracy is improved. The reconstruction compensation can update model parameters on line to cope with long-term changes of the system, intelligent distribution of compensation resources and self-adaptive adjustment of system changes are achieved, time-varying errors are effectively restrained, and precision, efficiency and robustness are taken into consideration.

The invention discloses a dynamic encryption transmission method and system for sound-light alarm signals, and relates to the field of data encryption, and the method comprises the steps: obtaining a dynamic environment entropy value; generating a pseudo-random chaotic sequence; obtaining a helium-oxygen mixed gas density parameter in the dynamic environment entropy value; segmenting a preset standard alarm waveform into a plurality of signal waveforms in a time domain, and generating a distorted signal waveform according to the helium-oxygen mixed gas density parameter; performing encryption modulation on the distorted signal waveform to generate an encrypted distorted signal waveform; determining a sending sequence and a sending interval of the encrypted distortion signal waveform by combining a pseudo-random chaos sequence based on an impedance fluctuation value and background acoustic spectrum characteristics in the dynamic environment entropy; generating a dynamic modulation signal based on the transmission sequence and the transmission interval; the dynamic modulation signal is sent to a water surface console through an umbilical cable; and receiving the guide pulse, performing phase conjugation processing on the guide pulse, generating an echo verification signal, and transmitting the echo verification signal back to the water surface console.

The invention discloses a six-dimensional force sensor dynamic decoupling and coupling processing method, and relates to the technical field of mechanical measurement sensing detection, and the method comprises the steps: building a dynamic decoupling relation of six-dimensional mechanical component and elastomer strain response coupling mapping based on an original voltage signal and a static calibration parameter of a six-dimensional force sensor; dynamically modulating the mechanical compensation factor of the dynamic decoupling relation through the environmental influence factor; according to the modulated mechanical compensation factor, performing real-time suppression and compensation on coupling deviation caused by environmental disturbance, and outputting a six-dimensional mechanical component subjected to environmental compensation; performing high-precision decoupling processing on the six-dimensional mechanical component subjected to environment compensation to generate a corrected six-dimensional mechanical component; performing closed-loop online updating on the mechanical compensation factor of the dynamic decoupling relation by using the corrected six-dimensional mechanical component; according to the invention, the coupling deviation caused by environmental disturbance is suppressed and compensated in real time, and the defects of the six-dimensional force sensor in the aspects of decoupling precision and measurement stability are improved.

The invention discloses a container damage detection method based on improved UPAD-YOLO, and relates to the technical field of container damage detection, and the method comprises the following steps: image collection: obtaining an image data set containing three types of defects of recess, broken hole and structure damage; preprocessing data, dividing a data set and labeling defect information; constructing a model, and embedding a dynamic modulation convolution module into the backbone network; feature enhancement: introducing a parallelization block perception attention module; feature fusion: integrating a self-adaptive spatial feature fusion module on a detection head; according to the method, multi-scale defect features can be effectively extracted, the detection precision and the anti-interference capability are improved, the problems that the multi-scale defect detection precision is limited and the detection effect is poor under a complex background are solved, and the method is suitable for container damage real-time detection in a large-scale logistics scene.

The invention discloses a vision-language-action robot control method and system based on task stage semantic enhancement, and the method comprises the steps: firstly constructing a task stage semantic system, and disassembling a task into stages with definite semantics; constructing a task stage observation module based on a pre-trained vision-language model, deducing a task state and abnormality in real time, and generating stage semantic features; modulation parameters are generated through a lightweight network, and dynamic modulation is carried out on the vision-language features to strengthen the vision feedback weight; and finally, fusing the modulated characteristics and the body state characteristics, and outputting a control instruction through an action generation network to form closed-loop control. According to the method, the multi-modal information weight can be dynamically balanced, task abnormity can be identified in real time, error correction is triggered, the false completion phenomenon is remarkably reduced, the task execution reliability, robustness and generalization ability of the robot in a complex dynamic environment are improved, and the method is suitable for various robot systems needing high-credibility operation.

The present application discloses and describes neurostimulation systems and methods that include, among other features, (i) neural stimulation through audio with dynamic modulation characteristics, (ii) audio content serving and creation based on modulation characteristics, (iii) extending audio tracks while avoiding audio discontinuities, and (iv) non-auditory neurostimulation and methods, including non-auditory neurostimulation for anesthesia recovery.

The invention relates to the technical field of emotion recognition, and discloses an emotion recognition method and system based on visual and auditory synergy, and the method comprises the steps: obtaining an original video stream and an original audio stream, introducing an explicit uncertainty quantification mechanism for a visual mode and an auditory mode, extracting features, and meanwhile, carrying out the recognition of the visual mode and the auditory mode. An uncertainty sequence that assesses its reliability is generated in parallel. Furthermore, the uncertainty sequence is used as a core guiding signal, and a cross-modal cross attention fusion process is dynamically modulated. The mechanism enables the model to adaptively suppress the contribution of an unreliable mode, and focuses on the interaction of high-quality signals, thereby effectively suppressing noise propagation, realizing robust multi-mode information complementation, and finally improving the accuracy of emotion recognition.

The invention discloses a newborn health monitoring system based on multi-modal data fusion, and belongs to the technical field of medical monitoring. According to the system, aiming at the problem of misjudgment caused by complex background interference and newborn development difference, a background perception feature decoupling enhancement network is constructed, a dual-channel architecture is matched with an adversarial loss function, foreground skin and background features are forcibly separated, and weak pathological features are enhanced in combination with a color space attention mechanism. Meanwhile, a development stage self-adaptive classification network is utilized, the gestational age serves as prior information to generate development codes, and dynamic modulation and weighted evaluation are conducted on fusion features through a premature infant and full-term infant double-branch classifier. According to the invention, the influence of environmental noise is effectively eliminated, and differentiated accurate health state recognition and risk early warning are realized.

The invention discloses a high-speed camera for semiconductor graph-free dark field detection and a detection method, and relates to the technical field of semiconductor detection, and the method comprises the steps: integrating an adjustable laser light source array, a high-performance TDI camera and a pre-scanning sensor, carrying out the brightness information collection of a pre-scanning region, and generating a pre-scanning reflectivity graph; sequentially irradiating a pre-scanning area with different illumination light through an adjustable laser light source array, generating a plurality of scanning images, and extracting features to generate a multi-modal differential feature field; predicting a main defect type in the region through a multi-mode illumination and defect feature collaborative enhancement algorithm to generate a corresponding illumination mode instruction; calculating an optimal TDI parameter through a TDI parameter dynamic modulation algorithm to generate a TDI camera control instruction; and scanning the wafer according to the two instructions, collecting original image data, inputting the original image data into the depth defect classification model, and outputting a defect information graph and corresponding information. The invention aims to solve the problem that semiconductor detection precision and efficiency are difficult to balance.

The invention discloses a system and method for generating, regulating and controlling a partially coherent space-time light field. The system comprises a pulse shaping assembly composed of a plurality of optical elements and an incident light source used for providing a completely coherent ultrafast pulse light field. An optical element in the pulse shaping assembly comprises a first blazed grating, a first cylindrical lens, a two-dimensional dynamic holographic modulator, a second cylindrical lens and a second blazed grating which are sequentially arranged in the light field propagation direction. The focal lengths of the first cylindrical lens and the second cylindrical lens are the same, and the distance between the adjacent optical elements is equal to the focal lengths of the first cylindrical lens and the second cylindrical lens; the two-dimensional dynamic holographic modulator is used for loading a specific complex amplitude modulation function to reconstruct the spatial-spectral characteristics of the light field. According to the invention, the two-dimensional space-spectrum function of the pulse is dynamically modulated, the generation of the partially coherent space-time wave packet is realized, and the amplitude and structure distribution of the space-time coherence can be designed and regulated as required.

The invention belongs to the technical field of production of cultural and creative products, and discloses a sorting and conveying device for production and preparation of cultural and creative products based on industrial vision. The device comprises a shading channel, a fixed support, a sliding frame, a motor, an intermittent transmission mechanism, a clamping mechanism, a synchronous belt, a polarization transmission shaft, an industrial detection camera and the like, the shading channel is a core detection space, the fixed support plays a role in supporting and leveling, the sliding frame can slide along a guide rail, and the motor drives an intermittent transmission turntable through a transmission part; the clamping mechanism adaptively clamps the ceramic product, and the optical system comprises a polarization disc, an inclined mirror surface and the like to realize accurate detection; 360-degree dead-corner-free detection of ceramic products, defect identification enhancement through dynamic polarization modulation, efficiency improvement through double-view-angle detection and self-adaptive clamping nondestructive protection can be achieved, through mechanical transmission precise control and optical dynamic modulation, the detection precision and efficiency are greatly improved, the labor cost is remarkably reduced, the method is suitable for large-scale production scenes, and intelligent upgrading of creative and creative production is assisted.

The invention discloses a high-fidelity visual character image generation method, system and device and a storage medium, which are corresponding schemes, in the scheme, spatial domain and spectral domain feature mutual learning is realized based on a double-domain font coding mechanism, and the problem of character structure disorder is solved; based on a frequency domain perception refining mechanism, a frequency domain signal is dynamically modulated to improve stroke detail precision, collaborative optimization of character details and background quality is realized through an instance-level scaling coefficient, and the bottlenecks of character distortion and background instability in the prior art are broken through. Besides, the method supports multi-language accurate generation, can be widely applied to commercial scenes such as advertisement design, product packaging, film and television posters, teaching materials and the like, and can greatly improve creation efficiency and reduce manual correction cost; and meanwhile, the feasibility of cooperatively improving the multi-modal generation efficiency by the cross-domain features is verified, a high-quality weak supervision training sample in a scarce scene can be provided for a character recognition model, the blank of related data is effectively filled up, and the method has technical advancement and market landing value.

The invention discloses a rapid sorting method and device for echelon utilization batteries and a storage medium, and relates to the technical field of echelon utilization batteries. In order to overcome the defects that an existing retired battery sorting method is inaccurate in sorting, low in efficiency, high in equipment requirement and incapable of achieving large-scale batch sorting, a battery sample is charged to preset cut-off voltage, and standing testing is conducted; recording a standing voltage sequence; carrying out feature extraction on the collected standing voltage sequence, and constructing a sorting feature vector required by battery sorting; dynamic modulation and gating weighted fusion are carried out on the feature vector obtained through wavelet decomposition and the cut-off voltage; classifying the performance grades of the batteries according to the SOH; and performing model training based on the modulation fusion features and the performance grades, obtaining feature vectors and cut-off voltages of the to-be-sorted echelon utilization batteries, and inputting the trained battery sorting model to obtain performance grade prediction results of the batteries. The method is mainly used for rapidly sorting large-scale retired batteries.

The invention discloses a millimeter wave radar gesture recognition method and system based on gating collaborative fusion, and belongs to the technical field of artificial intelligence and man-machine interaction. The method comprises the following steps: firstly, acquiring and preprocessing original data through a millimeter wave radar sensor; secondly, extracting a global spatial feature map (such as a two-dimensional point cloud map) capable of representing the overall motion track of the gesture and a plurality of local dynamic feature maps (such as a distance-time map, a Doppler-time map and the like) capable of representing dynamic information of the gesture in parallel; and then, the extracted multi-modal features are processed through the gated collaborative fusion network designed by the invention. The core of the method is that the global spatial feature map is processed through a first neural network and a global average pooling layer to generate a static global feature vector; dynamically modulating a local feature sequence processed by a second neural network through a gating unit by using the static global feature vector; and finally, fusing all modulated sequences, inputting the fused sequences into a time sequence processing module for analysis, and outputting a final gesture category. According to the method, through an innovative gating fusion mechanism, dynamic cooperation of multi-modal features is realized, the problem that a traditional fusion method cannot adapt to different gesture feature changes is effectively solved, and the accuracy and robustness of gesture recognition are remarkably improved.

The invention relates to a zero sample target counting device, and provides a technical scheme of fusing a fine-grained cross-modal enhancement module (FCE) and a multi-scale semantic adaptive module (MSA) in order to solve the problems of large counting error and poor generalization caused by insufficient cross-modal alignment precision and weak multi-scale modeling capability in a dense scene. A text-guided cross attention and self-adaptive gating mechanism is introduced through an FCE module, so that fine-grained semantic alignment of images and texts is realized, and the discrimination of visual features and the semantic perception capability are enhanced; multi-level visual features are fused through an MSA module, dynamic modulation is carried out in combination with text semantics, self-adaptive modeling of small targets and multi-scale targets is achieved, and high-precision target instance detection and number estimation are completed in the cross-modal decoding stage. The method is obviously superior to an existing model on multiple data sets without fine tuning, and the counting precision, robustness and open world generalization ability in a complex scene are improved.

The invention discloses a language model style backdoor defense method based on multi-level style representation, and the method achieves the comprehensive coverage of text style features and the efficient modeling of a high-dimensional space through the construction of a multi-level style representation frame and a self-adaptive multi-scale bandwidth selection method. A multi-dimensional suspicious degree scoring mechanism and a style-semantic balance repairing technology are adopted, and a sensitive neuron dynamic modulation and adaptive defense learning mechanism is combined, so that an evolvable defense system is constructed in a model, and three-dimensional defense from a sample level to a model level and from static protection to dynamic adaptation is realized; according to the technical means, the limitation of traditional single-dimensional representation is broken through, the accuracy and semantic fidelity of backdoor detection are remarkably improved, the defense process is made to have accuracy and practicability, innovativeness is high, resource consumption is low, the defense effect is continuous and stable, and complex and changeable text style backdoor attack scenes can be effectively dealt with.

The landslide image recognition method provided by the invention comprises the following steps: performing multiple times of residual processing on a target image to extract high-order and low-order semantic features; performing multi-scale feature extraction on the high-order features, dynamically modulating branch features through adaptive channel attention, and performing channel splicing to generate high-order fusion features; edge features are extracted from the high-order fusion features and the low-order features through depth separable convolution, height / width dimension aggregation is decoupled through a direction sensitive attention mechanism, direction weights are generated and cross enhancement is carried out, and high-order and low-order direction enhancement features are obtained; after the two are spliced, an output feature map is generated through two-way pooling aggregation and space attention enhancement, and a landslide division result is obtained through segmentation of a classifier. According to the method, the problems of insufficient feature fusion and weak boundary perception are solved, and the boundary precision and region consistency of landslide segmentation in a complex terrain are remarkably improved.

The invention provides a reconfigurable power amplifier based on dynamic modulation of a gate width adjustable transistor, and belongs to the technical field of radio frequency power amplifiers. The power amplifier is characterized in that a multi-mode grid width adjustable transistor is innovatively introduced to replace a traditional solidified structure parameter transistor to serve as a power amplifier tube; the transistor with the adjustable gate width is divided into a main unit cell and n auxiliary unit cells according to an interpolation index (or gate width), the main unit cell is responsible for determining the output power of the whole transistor, meanwhile, n small-power auxiliary unit cells are integrated to form an active load modulation module, and impedance matching optimization and linearity compensation are achieved by dynamically adjusting the on-off state of the active load modulation module. According to the structure, the linear expansibility of power output is guaranteed, a hardware basis is provided for micro impedance adjustment, cooperative regulation and control of system power efficiency dual modes and even linearity are achieved, and the power amplifier performance in a multi-band wireless communication system is effectively improved.