23results about How to "Accurate features" patented technology

The invention discloses a multi-attention mechanism network model and a thyroid ultrasound contrast focus segmentation method and system based on a multi-attention mechanism network. The multi-attention mechanism network model adopts an RFM-ResNeXt module to extract spatial and temporal feature information of an input video; aggregating the three-dimensional feature maps into a two-dimensional feature map through a bidirectional convolution LSTM module; inputting the two-dimensional feature map into a double attention module (DAM) to obtain an attention feature map; and carrying out convolution and up-sampling on the attention feature map to obtain a binary segmentation result image. The time-space information of the ultrasound contrast video can be effectively extracted by utilizing three attention mechanisms of channel attention, time attention and space attention, and the ablation area in the thyroid ultrasound contrast image can be automatically and accurately segmented.

The invention relates to a human body behavior recognition method, device and system based on WiFi and vision multi-mode fusion. The human body behavior recognition method processes WiFi channel state information and video data through two independent network models. The method comprises the following steps: firstly, carrying out feature extraction and recognition on WiFi channel state information by utilizing a WiFi deep learning model; then, performing feature extraction and recognition on the video data by using a visual deep learning model; and finally, splicing the identification results of the two and carrying out decision fusion. According to the method, multi-modal data fusion in a feature layer is avoided, and the influence of modal information change on a prediction result is reduced, so that the robustness of the human body behavior recognition method under the conditions of data noise, modal failure and the like and the accuracy of human body behavior recognition are improved.

The invention discloses a coarse cereal quality evaluation method and system based on a characteristic spectrum, and relates to the technical field of coarse cereal quality evaluation.The coarse cereal quality evaluation method comprises the following steps that light of a specific wave band is emitted to the surface of quinoa through a hyperspectral imager, and spectral data reflected by the surface of the quinoa are collected; the chenopodium quinoa willd surface spectrum change is captured through the spectrum extreme value attenuation index and the spectrum width index, the spectrum characteristic change coefficient is generated in combination with the machine learning model, and accurate judgment of mildewing of different degrees is achieved. Aiming at insufficient reflection caused by an oxide layer, the light emission duration is dynamically adjusted based on a Bayesian optimization algorithm, reliable spectral information is ensured to be obtained, and the adaptability and the detection precision are improved. In a complex environment, the system can stably and accurately identify mildew, delay is avoided, and high-quality data is provided for coarse cereal quality evaluation.

The invention discloses an improved YOLOv7 human body key point detection method. According to the method, manual labeling is carried out through LabelImage to obtain abundant training samples; replacing an original MP module in the YOLOv7 with an L-MP module to improve feature extraction, and adding a LiftDown Pool branch and residual connection; an EDH-Conv module is newly added between a Backbone module and a Head module of the YOLOv7 network, and the output of the EDH-Conv module is used as spatial weight distribution to correct an activated region, so that the detection capability of the network on a small target is further improved; an improved RSwin Transformer encoder is added to improve the detection effect of small targets and overlapped targets in a complex scene; a loss function is changed into an improved key point detection loss based on L2 loss by referring to Wise-IoU LOSS, and a dynamic focusing mechanism used for adjusting a key point error weight is introduced to accelerate convergence of the model. According to the method, the human body key point detection and identification capability in a complex scene is enhanced, and the method has higher robustness, accuracy and efficiency for human body key point detection, identification and counting tasks in sports.

The invention relates to the field of steel ladle automatic intelligent control, in particular to a steel ladle lining detection method and system. The steel ladle lining detection method comprises the following steps: acquiring point cloud image information of a steel ladle lining; discrete points in the point cloud image information are removed to obtain target point cloud data; performing edge detection on the target point cloud data to obtain a Canny edge image, and calculating a gray level co-occurrence matrix based on the Canny edge image to obtain texture features; extracting edge features through the Canny edge image, and fusing the edge features and the texture features to obtain a target feature vector; segmenting a damaged area at the target point cloud data by using a U-Net segmentation model, and outputting damage information through target feature vector classification; and performing three-dimensional reconstruction according to the target point cloud data and the damage information.

The invention relates to the field of intelligent automobile data processing, and discloses an automobile user behavior prediction method based on a data model, and the method comprises the following steps: 1, collecting user multi-modal data which comprises a user driving track, a purchase record, a vehicle-mounted video and a voice instruction; 2, carrying out standardization processing on the multi-modal data, and constructing a data matrix; 3, calculating a covariance matrix of the data matrix, and improving the stability of the covariance matrix through an optimization algorithm; step 4, eliminating noise features in the covariance matrix based on a random matrix theory, and extracting features with relatively high signal intensity; and 5, optimizing the cooperative relationship between the signal features by adopting a feature interaction modeling method. According to the method, the technical scheme of combining multi-modal data acquisition and standardized processing is adopted, and the fusion of high-dimensional and diversified data under the same analysis framework is realized by unifying the scale and distribution of different modal data.

The invention relates to the technical field of transformer vibration signal processing, in particular to a transformer vibration signal extraction method, system and equipment and a medium, and the method comprises the steps: collecting a vibration signal of a transformer in an operation process, and carrying out the preprocessing of the vibration signal; performing wavelet transform on the preprocessed vibration signal to obtain an energy time-frequency diagram of the vibration signal, and obtaining a wavelet energy-time distribution curve according to the energy time-frequency diagram; and determining an extreme point of the wavelet energy-time distribution curve, determining a starting point and an ending point of the vibration signal according to the set extreme point, and further extracting two periodic signal time domain waveforms in a power frequency period. According to the method, the time domain waveform of the transformer vibration signal can be efficiently and accurately obtained.

The present invention provides a method and device, equipment and storage medium for extracting water bodies in flood disasters, wherein the method comprises: obtaining a SAR image of a water body to be extracted, wherein the SAR image of the water body to be extracted comprises an image composed of a dual polarization band and a spectral characteristic band of a flood disaster SAR image; inputting the SAR image of the water body to be extracted into a preset flood extraction convolutional neural network to obtain a distribution map of water body information in flood disasters; wherein the preset flood extraction convolutional neural network is used to perform residual convolution and attention mechanism correction on the SAR image of the water body to be extracted, and perform hole convolution on the corrected SAR image of the water body to be extracted to obtain the distribution map of water body information in flood disasters. The present invention can achieve the purpose of extracting water body information based on SAR images of flood disasters with high precision and high accuracy, and effectively improves the extraction precision and accuracy of water bodies in flood disasters.

The invention belongs to the field of protein fluorescence intensity prediction, and particularly relates to a protein sequence fluorescence intensity prediction method, system and device and a storage medium, and the method comprises the steps: carrying out the feature extraction of a training data set, combining random noise, training a generator and a discriminator, obtaining a generative adversarial model, supplementing data through the generative adversarial model, and obtaining a protein sequence fluorescence intensity prediction result. The method comprises the steps of obtaining a prediction data set, then extracting semantic features of the prediction data set, performing local mode enhancement and fusion on hidden layer features to obtain final features, inputting the final features into a prediction network to train the prediction network to obtain an optimal prediction network, and inputting an actual protein sequence into the optimal prediction network. The fluorescence intensity corresponding to the actual protein sequence is obtained. The method has the effect of improving the accuracy of predicting the fluorescence intensity of the protein sequence.

The application discloses a visual information recognition method and system based on eye movement attention, a medium and equipment, and relates to the technical field of visual analysis. Eye movement data is obtained through an eye jump experiment, and after processing, the eye movement data first enters a multi-head attention unit to calculate the self-attention weight of the eye movement data to perform feature selection in the first stage, then enters a local attention unit to further extract local attention features and obtain key information of the eye movement data, and finally, the data information is added through an adding unit, so that the extracted features are more comprehensive and more accurate. In the case of limited eye movement data, the key information in the features is maximized to more accurately detect visual information.

The application discloses an intention recognition method, device, equipment and medium. In the embodiment of the application, after a text matrix corresponding to a control text is input into an intention recognition model, the intention recognition model performs feature extraction operations on the text matrix based on a preset number of feature extraction methods, and a preset number of feature matrices are obtained. Compared with the prior art in which only one feature extraction method is used to perform a feature extraction operation on a text matrix, the features extracted based on the preset number of feature extraction methods are more accurate, which effectively improves the accuracy of the intention recognition model in identifying an intention and improves the user experience.

The invention provides an audio processing integration method and device of a hearing aid. The method comprises the following steps: acquiring an audio signal acquired by a microphone of the hearing aid in real time; processing the audio signal through a feature extraction network to obtain first audio feature data; performing frequency domain conversion, noise distribution prediction and weighted filtering processing on the first audio feature data based on a noise suppression network to obtain a first noise reduction audio feature; processing the first audio feature data by using an adaptive enhancement network, and obtaining a second enhanced audio feature by randomly selecting a frequency reference point and collecting and processing a feature point; fusing the first noise reduction audio feature and the second enhancement audio feature through a cross attention network to obtain a fused audio feature; and finally, processing the fused audio features by using an audio enhancement network to obtain an enhanced audio signal and outputting the enhanced audio signal. According to the invention, the voice definition and sound quality performance of the hearing aid in a complex environment can be improved, and the hearing experience of a hearing impaired user is improved.

Embodiments of the present application relate to a feature extraction method and device, electronic equipment and storage medium. The method comprises: extracting a feature map of a target image; inputting the feature map into a trained channel attention feature extraction model and a spatial attention feature extraction model respectively to obtain channel attention features and spatial attention features of the target image; using a preset weighted distance feature fusion algorithm and a weighted product feature fusion algorithm to perform feature fusion on the channel attention features and the spatial attention features respectively to obtain weighted distance fusion features and weighted product fusion features; determining a self-attention parameter based on the weighted distance fusion features and the weighted product fusion features; and determining self-attention features of the target image based on the feature map and the self-attention parameter. Thus, the accuracy of feature extraction is improved.

The present invention discloses a forest fire image detection method based on a generative adversarial network, and the steps are as follows: step 1: obtain forest fire monitoring images to form a data set; step 2: extract the image feature matrix of each image in the data set; step 3: construct the dictionary and classifier parameters in the discriminator and generator through deep dictionary learning; step 4: use the image feature matrix as the input of the discriminator, obtain the dictionary and sparse matrix of each layer in the discriminator; obtain the dictionary and generated image of each layer in the generator; step 5: according to the objective function, repeat step 4 until the dictionary, sparse matrix and generated image are stable; step 6: use the stabilized generative adversarial network to detect forest fire. The present invention constructs the dictionary and classifier parameters in the discriminator and generator of the generative adversarial network through deep dictionary learning, and the constructed generative adversarial network can more effectively identify anomalies, making anomaly detection more accurate.

The embodiments of this specification provide a network training method and device for event detection. It includes a dynamic perception network and a target feedback network. The dynamic perception network includes a plurality of alternative computing layers arranged in sequence. When training the dynamic perception network, for any word in the acquired sample text, the first several computing layers are gradually selected from the dynamic perception network, and the intermediate feature vector of the word is determined by the first several computing layers and the initial feature vector of the word, and the first predicted label is determined based on the first classifier and the intermediate feature vector of the word. When the first difference between the first predicted label and the corresponding annotation label is less than the first preset threshold, the first several computing layers are used as the computing layers corresponding to the word, and the intermediate feature vector of the word is used as the first feature vector of the word. The first prediction loss is determined based on the first difference, and the model parameters in the computing layer corresponding to the word are adjusted in the direction of reducing the first prediction loss.

The invention discloses a railway roadbed disease rapid diagnosis method based on spectrogram analysis, and relates to the technical field of signal processing and track engineering. The method comprises the following steps: firstly, acquiring acceleration signal samples influenced by typical roadbed diseases by using acceleration sensors mounted at different positions of a vehicle and a roadbed; secondly, performing multi-source information fusion by adopting DS data fusion; combining different frequency scale features to construct a disease feature vector, and using a spectrogram theory to form a railway roadbed disease spectrogram; and then, carrying out spectral analysis on the acceleration signal monitored in real time, and comparing with a disease spectrum library to judge the roadbed disease. The method has the beneficial effects that the health condition of the railroad bed can be reflected from different angles and levels through the fused signal source, the limitation of a single detection means is overcome, and the comprehensiveness of disease detection is improved; the disease degree and position can be accurately evaluated by using a spectrogram theory and a positioning system, the railway roadbed disease diagnosis efficiency is remarkably improved, and the monitoring cost is effectively reduced.

The present application provides a method, device, electronic device and computer-readable storage medium for extracting features from a multispectral image. The method comprises: obtaining a multispectral image to be processed, and synthesizing the multispectral image by bands to obtain at least two synthesized images, wherein the number of bands included in the synthesized image is less than the number of bands included in the multispectral image; extracting features from each synthesized image to obtain a first feature map corresponding to each synthesized image; recombining features in each first feature map to obtain a second feature map corresponding to each first feature map; and splicing each second feature map to obtain a spliced ​​feature map corresponding to the multispectral image. Through the present application, the feature extraction capability of multispectral images can be improved.

The invention provides a dynamic graph data feature extraction method, device and equipment based on a lightweight dynamic graph neural network. The method comprises the following steps: converting dynamic data into dynamic graph structure data; constructing a loss function; performing convolution operation on the dynamic graph structure data to extract dynamic graph data features; feature transformation operation and nonlinear activation operation in convolution operation are removed, and dynamic graph data features are extracted; judging whether the loss value of the loss function is converged or not or whether the number of iterations reaches an upper line or not; if yes, outputting dynamic graph data features; and if not, carrying out iterative optimization on the loss function by utilizing a stochastic gradient descent method. According to the method, by reducing the feature transformation operation and the nonlinear activation operation, the calculation complexity is remarkably reduced, so that the calculation complexity is reduced, and the training efficiency of the model is improved. In addition, by introducing the symmetrically standardized adjacent tensor, the description capability of the model on the cross-space-time association relationship in the dynamic graph data is enhanced, so that the extracted features are more accurate.

The present invention provides a two-stage sparse laser radar point cloud feature extraction method based on cells, which comprises: step 1: reading original three-dimensional point cloud data; step 2: dividing a frame of point cloud into N*S cells; step 3: marking the cells corresponding to each point cloud; step 4, traversing each cell, and entering step 5 when the number of points in the cell is greater than a set value; step 5, calculating the normalized mean coordinates of the cell; step 6, constructing the covariance matrix of the cell and solving it to obtain the eigenvalue; step 7, judging the planarity or linearity of the cell according to the eigenvalue; step 81, the cell is judged to be planar, the covariance of each point in the cell is calculated, the eigenvalue is obtained, and step 91 is entered; step 82, the cell is judged to be linear, the covariance of each point in the cell is calculated, the eigenvalue is obtained, and step 92 is entered; step 91, according to the planarity of the point cloud, the plane feature points are obtained; step 92, according to the linearity of the point cloud, the line feature points are obtained.

The invention provides a video feature vector extraction method, and belongs to the technical field of video processing. Comprising the following steps: splitting a video into video segments by using a segment detection mode; acquiring a key frame of each video clip; obtaining a feature vector of each key frame by using an image element analysis method based on an image HSV color space; and taking a set of the feature vectors of all the key frames as the feature vector of the video. According to the method, the feature vector of the video can be extracted, the video features are stored in the specific database, and when video retrieval needs to be carried out, similarity calculation is carried out by utilizing the features of the query image or the video and the features in the database, so that the required video clip can be queried; according to the image feature vector extracted by the method, color, texture, spatial distribution and shape information of the image can be considered, and the features of the image can be reflected more accurately.

The application belongs to the technical field of transcranial magnetic stimulation, and particularly relates to a transcranial magnetic stimulation electroencephalogram positioning method based on multi-feature fusion; the method comprises the following steps: acquiring electroencephalogram data of a subject in real time during a magnetic stimulation process; intercepting and preprocessing the acquired electroencephalogram data, calculating power spectrum density, activity index and permutation entropy of each lead in each frequency band; fusing the above features, calculating comprehensive indexes of each lead; performing feature processing on the comprehensive indexes of each lead, and calculating final features of each lead; performing descending order sorting on the final features of each lead, and selecting three leads with high rankings; solving geometric position centers of the selected three leads, and obtaining position information of a magnetic stimulation point. In the transcranial magnetic stimulation scene, the application can determine the position of transcranial magnetic stimulation without the aid of MRI images, through real-time electroencephalogram feature analysis, so as to reduce positioning cost and improve timeliness.

The utility model discloses an automatic sampling device for raw materials, which relates to the technical field of material sampling and comprises an incoming material conveying belt, a sampling station, an automatic sampling device and a sampling device. A transfer belt is arranged at the tail end of the sampling station and communicated with the sample preparation system; a transfer conveying belt is arranged on one side of the sampling station and is communicated with the material returning field; sampling grooves are formed in the bottoms of the sampling stations. The device is matched with a system sampling method to realize automation of a series of links from raw material conveying to sampling, transferring, sample preparation and the like; in the whole process, the influence of human factors on a sampling result is reduced, system errors caused by manual misoperation or manual intervention are avoided, the fairness and objectivity of the sampling result are improved, the labor intensity of workers is reduced, the working efficiency is improved, and the labor intensity of the workers is reduced. The method ensures that the taken samples can more accurately represent the overall characteristics of bulk materials, and provides a more reliable basis for trade settlement.