Abstract: An online detection device underwater elements includes an LIBS system in a sealing pressure chamber and an external airflow control system. The airflow control system has a gas probe bin and a gas source. An opening is formed at one end of the gas probe bin while the other end and the sealing pressure chamber are hermetically partitioned through a glass window. A laser in the LIES system outputs laser to an underwater object surface to be detected for generating plasma spectra. A spectrometer collects plasma spectra returned along an original optical path. When the device operates in water, the balance gas storage tank produces gas with the same pressure as underwater. A flow model is invoked according to the current water pressure to accurately control the air flow rate to form a stable gas environment in the gas probe, which improves the plasma excitation and collection efficiency.

Abstract: A new image formation model for descattering different from the previous works based on atmospheric scattering model is established. It can provide a corresponding physical explanation for each scattering environment, illustrate a relationship between a light source position and scattering conditions and solve the problem of scattering removal under non-uniform illumination. Firstly, a dark angle of a camera is removed by an integrating sphere; then, a pure scattering part is found by an improved dark channel-like prior method; and finally, a zenith angle, an azimuth angle, an extinction coefficient and other related physical quantities are calculated according to the light attenuation and backscattering distribution of a scattering medium to eliminate the backscattering and light attenuation. Theoretically, the model can be used in any uniform scattering medium, such as underwater and fog environments, and works well in high-density media.

Abstract: Disclosed are a method and a system for mental state perception and a computer-readable storage medium. The method includes: acquiring image sequences with timestamps and millimeter-wave radar raw data with timestamps; preprocessing the image sequences and the millimeter-wave radar raw data; analyzing the head region image sequences to obtain head vibration signal features; calculating face region image sequences obtained from preprocessing by using a remote photovolumetric pulse wave recording method to obtain a first heart rate; analyzing the original millimeter-wave radar data sequence to obtain a second heart rate and a breathing rate; fusing the first heart rate, the second heart rate and the breathing rate to obtain a fused heart rate and breathing rate; performing feature extraction on facial change information by a Transformer-like network; establishing a non-contact multi-modal mental perception model for prediction to obtain a predicted result of the mental state.

Abstract: A feature extraction system, method and apparatus based on neural network optimization by gradient filtering is provided. The feature extraction method includes: acquiring, by an information acquisition device, input information; constructing, by a feature extraction device, different feature extraction networks, performing iterative training on the networks in combination with corresponding training task queues to obtain optimized feature extraction networks for different input information, and calling a corresponding optimized feature extraction network to perform feature extraction according to a class of the input information; performing, by an online updating device, online updating of the networks; and outputting, by a feature output device, a feature of the input information. The new feature extraction system, method and apparatus avoids the problem of catastrophic forgetting of the artificial neural network in continuous tasks, and achieves high accuracy and precision in continuous feature extraction.

Abstract: The present disclosure belongs to the field of defect detections and discloses a method for constructing a defect detection model, a method for detecting a defect and a related apparatus, obtaining an initial training image, and adding a simulated anomaly to the initial training image to obtain a simulated anomaly training image; training a preset defect recognition model according to the initial training image and the simulated anomaly training image to obtain defect position information and mask prompt information; training a preset defect segmentation model according to the defect position information and the mask prompt information; and fusing the trained defect recognition model and defect segmentation model to obtain a defect detection model; the defect recognition model includes a teacher network branch, a student network branch and an autoencoder network branch; and an output difference between the teacher network branch and the student network branch is the defect position information.

Abstract: Provided are a device, a system and a method for acquiring a force information based on a bionic structure, including: a force information acquisition layer and a magnetic field signal acquisition chip; wherein a permanent magnet is embedded in the force information acquisition layer; wherein the force information acquisition layer has an elastic structure configured to generate a deformation corresponding to a first force information of a force after being subjected to the force, so that the permanent magnet moves with the deformation to generate a magnetic field signal corresponding to the force information; wherein the magnetic field signal acquisition chip is arranged in parallel with the force information acquisition layer, and is configured to acquire the magnetic field signal and convert the magnetic field signal into an electrical signal.

Abstract: A system for reconstructing a magnetic particle image based on a pre-trained model aims to address the influence by point spread function and reduce the computational and time costs, which results in low reconstruction accuracy, or high acquisition time and computational costs for high-precision images. The system is implemented by: generating a simulation system matrix; pre-training a pre-constructed neural network model, and fine-tuning a pre-trained neural network model by performing a downstream task; and inputting real data corresponding to the downstream task into the pre-trained neural network model after fine-tuning, thereby playing an auxiliary role to acquire a high-quality reconstructed MPI image. The system fits the relationship between different harmonics, which helps enhance frequency-domain information.

Abstract: Disclosed is a method for updating a node model that resists discrimination propagation in federated learning. The method includes: obtaining a node model corresponding to a data node; calculating a mean value of the distribution of class features and a quantity ratio corresponding to training data of the data node, calculating a distribution weighted aggregation model based on the node model, the mean value of the distribution of class features and the quantity ratio; calculating a regularization term corresponding to the data node based on the node model and the distribution weighted aggregation model; calculating a variance of the distribution of the class features corresponding to the data node, calculating a class balanced complementary term by using a cross-domain feature generator; and updating the node model based on the distribution weighted aggregation model, the regularization term, and the class balanced complementary term.

Abstract: Disclosed is a disentangled personalized federated learning method via consensus representation extraction and diversity propagation provided by embodiments of the present application. The method includes: receiving, by a current node, local consensus representation extraction models and unique representation extraction models corresponding to other nodes, respectively; extracting, by the current node, the representations of the data of the current node by using the unique representation extraction models of other nodes respectively, and calculating first mutual information between different sets of representation distributions, determining similarity of the data distributions between the nodes based on the size of the first mutual information, and determining aggregation weights corresponding to the other nodes based on the first mutual information; the current node obtains the global consensus representation aggregation model corresponding to the current node.

Abstract: A system for reconstructing a magnetic particle image based on adaptive optimization of regularization terms includes: a MPI device for scanning an imaging object to acquire a voltage response signal; a signal processor for constructing a system matrix; and a control processor for reconstructing the magnetic particle image based on an arbitrarily selected regularization term, inputting the reconstructed magnetic particle image to a regularization-term adaptive optimization neural network model for enhancement processing, taking the enhanced magnetic particle image as a first image, and calculating a loss value between the first image and an initial image to acquire a final reconstructed magnetic particle image. The system adopts a neural network model-based automatic learning approach, instead of the approach of manually selecting regularization terms and adjusting parameters, to improve the reconstruction efficiency and quality of the magnetic particle image.

Abstract: Disclosed is a method for updating a node model that resists discrimination propagation in federated learning. The method includes: obtaining a node model corresponding to a data node; calculating a mean value of the distribution of class features and a quantity ratio corresponding to training data of the data node, calculating a distribution weighted aggregation model based on the node model, the mean value of the distribution of class features and the quantity ratio; calculating a regularization term corresponding to the data node based on the node model and the distribution weighted aggregation model; calculating a variance of the distribution of the class features corresponding to the data node, calculating a class balanced complementary term by using a cross-domain feature generator; and updating the node model based on the distribution weighted aggregation model, the regularization term, and the class balanced complementary term.

Abstract: Disclosed is a disentangled personalized federated learning method via consensus representation extraction and diversity propagation provided by embodiments of the present application. The method includes: receiving, by a current node, local consensus representation extraction models and unique representation extraction models corresponding to other nodes, respectively; extracting, by the current node, the representations of the data of the current node by using the unique representation extraction models of other nodes respectively, and calculating first mutual information between different sets of representation distributions, determining similarity of the data distributions between the nodes based on the size of the first mutual information, and determining aggregation weights corresponding to the other nodes based on the first mutual information; the current node obtains the global consensus representation aggregation model corresponding to the current node.

Abstract: Disclosed is a dialogue emotion correction method based on a graph neural network, including: extracting acoustic features, text features, and image features from a video file to fuse them into multi-modal features; obtaining an emotion prediction result of each sentence of a dialogue in the video file by using the multi-modal features; fusing the emotion prediction result of each sentence with interaction information between talkers in the video file to obtain interaction information fused emotion features; combining, on the basis of the interaction information fused emotion features, with context-dependence relationship in the dialogue to obtain time-series information fused emotion features; correcting, by using the time-series information fused emotion features, the emotion prediction result of each sentence that is obtained previously as to obtain a more accurate emotion recognition result.

Abstract: An industrial 5G dynamic multi-priority multi-access method based on deep reinforcement learning includes the following steps: establishing an industrial 5G network model; establishing a dynamic multi-priority multi-channel access neural network model based on deep reinforcement learning; collecting state, action and reward information of multiple time slots of all industrial 5G terminals in the industrial 5G network as training data; training the neural network model by using the collected data until the packet loss ratio and end-to-end latency meet industrial communication requirements; collecting the state information of all the industrial 5G terminals in the industrial 5G network at the current time slot as the input of the neural network model; conducting multi-priority channel allocation; and conducting multi-access by the industrial 5G terminals according to a channel allocation result.

Abstract: A second near-infrared window/first near-infrared window dual-mode fluorescence tomography system having a lighting module, an excitation module, a second near-infrared window collection module, a first near-infrared window collection module, a CT imaging module and a central control module. The central control module is configured to reconstruct second near-infrared window three-dimensional and tomographic images and first near-infrared window three-dimensional and tomographic images based on the white light images, the second near-infrared window fluorescent images, the first near-infrared window fluorescent images, and the CT images. The reconstructed three-dimensional space tumor signal has depth characteristics, which is closer to the real distribution of tumors, such that the reconstruction position is more accurate. The three-dimensional shape of the tumor is displayed intuitively and clearly at any angle with the usage of image display unit.

Abstract: Disclosed are a target speaker separation system, an electronic device and a storage medium. The system includes: first, performing, jointly unified modeling on a plurality of cues based a masked pre-training strategy, to boost the inference capability of a model for missing cues and enhance the representation accuracy of disturbed cues; and second, constructing a hierarchical cue modulation module. A spatial cue is introduced into a primary cue modulation module for directional enhancement of a speech of a speaker; in an intermediate cue modulation module, the speech of the speaker is enhanced on the basis of temporal coherence of a dynamic cue and an auditory signal component; a steady-state cue is introduced into an advanced cue modulation module for selective filtering; and finally, the supervised learning capability of simulation data and the unsupervised learning effect of real mixed data are sufficiently utilized.

Abstract: Disclosed is an automatic depression detection method using audio-video, including: acquiring original data containing two modalities of long-term audio file and long-term video file from an audio-video file; dividing the long-term audio file into several audio segments, and meanwhile dividing the long-term video file into a plurality of video segments; inputting each audio segment/each video segment into an audio feature extraction network/a video feature extraction network to obtain in-depth audio features/in-depth video features; calculating the in-depth audio features and the in-depth video features by using multi-head attention mechanism so as to obtain attention audio features and attention video features; aggregating the attention audio features and the attention video features into audio-video features; and inputting the audio-video features into a decision network to predict a depression level of an individual in the audio-video file.

Abstract: A bi-directional interaction network (BINet)-based person search method, system, and apparatus are provided. The method includes: obtaining, as an input image, a tth frame of image in an input video; and normalizing the input image, and obtaining a search result of a to-be-searched target person by using a pre-trained person search model, where the person search model is constructed based on a residual network, and a new classification layer is added to a classification and regression layer of the residual network to obtain an identity classification probability of the target person. The method improves the accuracy of the person search.

Abstract: A neurovascular coupling analytical method based on an electroencephalogram and functional near-infrared spectroscopy includes: S100: acquiring an electroencephalogram signal and a brain hemodynamic signal; S110: extracting an event-related potential signal from the electroencephalogram signal; S120: extracting a time characteristic from the event-related potential signal; S130: extracting a hemodynamic response function from the brain hemodynamic signal; S140: extracting an amplitude characteristic and time characteristics from the hemodynamic response function; and S150: analyzing influence of the time characteristic of the event-related potential signal on the amplitude characteristic and the time characteristics of the hemodynamic response function to obtain a coupling result. The time characteristic of the event-related potential signal is a delay.

Abstract: The present invention relates to ground support equipment for aerospace engineering, and particularly relates to an assembly and test operation robot for a space station experimental cabinet. The assembly and test operation robot comprises a mobile lifting platform, a comprehensive monitoring system, a rotating clamping mechanism, a multifunctional adapter and a science experimental cabinetet, wherein the mobile lifting platform is used for regulating the horizontal position and the height position of the science experimental cabinetet to realize assembly and transportation functions of the experimental cabinet; the rotating clamping mechanism is installed on the mobile lifting platform to realize clamping and rotation functions of the science experimental cabinet; the multifunctional adapter is installed on the rotating clamping mechanism to carry the science experimental cabinet; and the comprehensive monitoring system is used to monitor the assembly state of the science experimental cabinet in real time.