10results about How to "High computational complexity" patented technology

This invention discloses a fall detection method, apparatus, device, and medium. The method includes: determining target noise point detection information corresponding to the target radar detection information of a target person based on personnel radar detection information detected by millimeter-wave radar and noise point detection information detected by an acoustic detection device; determining the detection mode of the target person based on the target radar detection information, and determining target detection information from the target radar detection information and target noise point detection information based on the detection mode; wherein the detection mode includes an obstruction mode, a stationary mode, and a motion mode; and detecting the fall state of the target person based on the target detection information. This invention uses a combination of millimeter-wave radar and acoustic detection for fall detection, enabling continuous fall state detection even when the target person is obstructed by objects indoors, improving the accuracy and efficiency of fall detection; and since both millimeter waves and acoustic waves are invisible waves, user privacy can be effectively protected.

The invention provides a polarization image defogging method based on a neural network, and belongs to the technical field of image processing, and the method comprises the steps: obtaining a plurality of fogged polarization images, and generating a polarization image package; constructing a neural network defogging model; and based on the constructed neural network defogging model, carrying out defogging processing on the foggy polarization image in the generated polarization image packet to obtain a fogless image. According to the image defogging method, the deep learning and the polarization imaging technology are combined, the problem that the calculation complexity of the polarization imaging technology is high can be effectively solved, the overall generalization ability and defogging level are improved, and better robustness and high adaptability can be shown in the face of different types of fog and different degrees of haze.

The application relates to the technical field of direction of arrival estimation, and specifically discloses a cross-platform complex domain feature enhancement coherent super-resolution DOA estimation method, which comprises the following steps: firstly, a complex domain CVSIMO learning model is constructed; sampling data is preprocessed; forward propagation of the CVSIMO learning model is carried out; backward propagation and parameter optimization of the CVSIMO learning model are carried out; data reconstruction and signal separation are carried out; cross-platform super-resolution DOA estimation is carried out; and finally, model simulation experiments are carried out. The cross-platform complex domain feature enhancement coherent super-resolution DOA estimation method can realize feature mining of multiple point source signals, converts a multi-source estimation problem into a single point source DOA estimation problem by using a complex domain neural network model, enhances the features of coherent signals by using the advantages of the complex domain neural network, improves the performance and precision of the super-resolution DOA estimation algorithm, can also realize data separation of multiple coherent signals and cross-platform DOA estimation, and realizes real angle estimation through ingenious feature solving.

The application discloses a load aggregation body prediction method and system of an integrated framework, and the method comprises the following steps: applying a prediction result of a trained load aggregation body prediction model to perform scheduling; wherein the construction of the load aggregation body prediction model comprises the following steps: obtaining load aggregation body historical data; performing external influence factor analysis on the load aggregation body historical data, and screening important external influence factors; applying a K-shape algorithm to cluster the load aggregation body historical data, and dividing the load aggregation body historical data into a plurality of clusters; constructing a load aggregation body prediction model, including a CNN-LSTM-ATTENTION prediction model and a GBDT prediction model; performing load prediction on each cluster through the CNN-LSTM-ATTENTION prediction model by using the screened important external influence factors and the divided plurality of clusters; and applying the GBDT prediction model to perform integrated prediction on the load prediction results of the clusters, thereby obtaining a load aggregation body prediction result; and the application can improve the prediction accuracy of the load aggregation body prediction model.

The application discloses a general information transmission optimization method, comprising the following steps: setting the number of satellite communication sites as N, and the time interval of message transmission between the satellite communication sites as 1 minute; making the N satellite communication sites perform forward transmission in the 0th minute to the (N / 2)-1th or ((N-1) / 2)-1th minute of the message transmission; making the N satellite communication sites perform reverse reverse sequence transmission in the (N / 2)-1th or ((N-1) / 2)-1th minute to the (N / 2)+[(N / 2)-1]th or ((N-1) / 2)+[(N-1) / 2)-1]th minute of the message transmission; obtaining a signal transmission model under the cyclic recursive transmission of the forward information transmission and the reverse reverse sequence information transmission, and optimizing the information transmission based on the signal transmission model. The application has high realization, is easy to understand, can simplify complex problems, and can be realized through programming, so that the calculation complexity is greatly improved.

The application provides a single-pixel imaging multi-target positioning method based on multi-direction projection optimization, and relates to the field of moving target positioning and tracking, which comprises the following steps: based on prior position information of multiple targets, a Hough transform is used to extract a candidate projection angle set of collinear targets, and clustering analysis is performed on multiple candidate projection angles to obtain a key projection angle set; by using the key projection angle set, a Fourier central slice theorem is used to obtain one-dimensional projection curves corresponding to each key projection angle in the corresponding direction in the target scene; according to the one-dimensional projection curves, projection strips of the targets in the corresponding direction are extracted, combined strips are generated through spatial mapping and splicing, and an image containing a target region is obtained; after morphological filtering processing is performed on the image, position estimation is performed by using a centroid coordinate, and the position positioning of multiple targets is obtained. The application can effectively reduce the projection angle, reduce the measurement times and improve the positioning efficiency under the premise of ensuring the positioning accuracy.

The application discloses a decision tree binning calculation optimization method, equipment, medium and computer program product. The decision tree binning calculation optimization method comprises the following steps: obtaining memory storage data corresponding to a target decision tree and an index table corresponding to the memory storage data, wherein the index table at least comprises a node binning label corresponding to each decision tree binning; then, based on the size of each decision tree binning, the node binning label is segmented to obtain a segmented batch of each binning label; then, based on the index table, each decision tree binning batch data is extracted; then, based on the decision tree binning size information corresponding to each segmented batch of binning label, a corresponding number of parallel computing threads is matched for each decision tree binning batch data; then, based on the number of parallel computing threads, parallel decision tree binning calculation is performed on each decision tree binning batch data to obtain a target decision tree binning calculation result. The application solves the technical problem of low calculation efficiency during decision tree binning.

This invention provides a Fourier-Melin transform image geometric matching method and system with decoupled logarithmic domain, comprising: estimating the rotation angle using Fourier-Melin transform; rotating the image to be matched in the current frame relative to the template image in the previous frame to obtain an axially aligned image search region; transforming the template image and the image to be matched from Cartesian coordinates to logarithmic polar coordinates to obtain the image search region corresponding to the logarithmic polar coordinates; sequentially segmenting the image search regions corresponding to the image to be matched and the template image along the θ-axis according to a set ratio; stitching the horizontally correlated and vertically correlated image regions into two twin image search regions; calculating the cross power spectrum through phase correlation to obtain the scale factors in the horizontal and vertical directions; and aligning the images to be matched according to the scale factors in the two directions. This invention has the ability to estimate scale and rotation proportionally, and is scale adaptive, and is robust to out-of-plane flipping and rotation of the target object.

The invention discloses a robust time difference of arrival positioning method in a mixed line-of-sight / non-line-of-sight scene, and the method comprises the steps: constructing a joint optimization model which comprises a target position, a sparse NLOS error vector and a reference station common deviation term, and carrying out the iteration solving through employing a block coordinate descent (BCD) strategy; the sparsity of an NLOS error is described through L1 regularization or logarithm regularization, and position estimation is efficiently updated in combination with a Levenberg-Marquardt algorithm, so that low-complexity real-time positioning is realized; according to the method, an external convex optimization solver is not needed, and the calculation time is only milliseconds; the system, the medium and the electronic equipment are used for realizing the robust time difference of arrival positioning method under the mixed line-of-sight / non-line-of-sight scene. The positioning precision and robustness in a complex propagation environment, especially when the reference station is influenced by NLOS, are significantly improved, and the defect that modeling of the scene is insufficient in a traditional method is overcome.