35results about How to "Localized" patented technology

The invention discloses a face recognition method based on weighted diagnostic sparseness constraint nonnegative matrix decomposition, and mainly aims to solve the problem that the method in the prior art is not robust to an obscured face and is of low recognition rate. According to the technical scheme, the method comprises the following steps: (1) constructing a nonnegative weight matrix according to the obscured area of a test image; (2) introducing the weight matrix into a general KL divergence objective function, applying a sparseness constraint to a basis matrix, and applying intra-class and inter-class divergence constraints to a coefficient matrix to get a weighted diagnostic sparseness constraint nonnegative matrix decomposition objective function; (3) solving the objective function, and decomposition-training a data matrix to get a basis matrix and a coefficient matrix; (4) projecting a test data matrix on the basis matrix to get a corresponding low-dimensional representation set, and taking the low-dimensional representation set as final test data; and (5) using a nearest neighbor classifier to classify the test data by taking the coefficient matrix as training data, and outputting the result. By using the method, the effect of obscured face recognition is improved. The method can be used in identity recognition and information security.

This invention provides a network routing method used in providing optimum route for transmitting service flows between a source node and a destination note including: a, responding to a global routing request sent by the source note to find out at least one route meeting the request of bandwidth for transmitting service flows with the destination node, b, selecting a route with the optimum delay and stability in at least one route as the optimum route and c, preserving resources for the optimum route for transmitting service flows. This invention also provides a network routing device.

A digital watermark method resisting against geometric attack includes relating watermark information to image character region, generating local character region with invariance on geometric attack by utilizing most stable Harris character point in image, making said region to be round character region by image normalization, carrying out two-stages of wavelet transforms on these round character regions, selecting revision coefficient regions to embed watermark at the second stage wavelet transform and realizing watermark-embedding. The method for picking up watermark is also disclosed.

The invention relates to a wireless service flow prediction method based on weighted federated learning. The method comprises the following steps: a control center pushes a plurality of pre-training models to a base station side; the base station side performs model training according to the local data and transmits the trained model to a control center; the control center fuses the models according to a weighting rule and feeds back the fused models to the base station, and more weights are given to the local model in the weighting rule; and the base station predicts the wireless service flowat the future moment according to the obtained final model. According to the wireless service flow prediction method provided by the invention, on the basis of a control center model aggregation strategy, the weighted aggregation rule is used for replacing an average strategy, so that the phenomenon of inaccurate prediction caused by data isomerism can be avoided, and the overall prediction precision of distributed wireless service flow prediction is improved.

The invention discloses an incidence relation excavation method for a text-oriented knowledge unit. The method comprises the following steps of: (1) carrying out aggregation for a text set, finding a text subset with a similar theme, on the base, utilizing nonsymmetry of term distribution in the text and excavating a linear incidence relation between texts; (2) utilizing locality of the incidence relation of a knowledge unit pair, and generating a candidate knowledge unit pair; (3) based on characteristics of term word frequency, distance and semantic types of the knowledge unit pair, carrying out a bi-level classification for the candidate knowledge unit pair, and distinguishing the incidence relation of the knowledge unit pair. In the incidence relation excavation method, numbers of candidate knowledge units can be greatly reduced, and time complexity of relation excavation can be effectively reduced on the premise that accuracy is ensured.

The invention discloses a sparse local decomposition and restructuring algorithm of joint motion sensing; the method includes steps of 1), calculating length of sides in a triangular mesh and size of dihedral angles corresponding to sides (the two values are recorded as LA); 2), calculating residual difference of LA vector in every triangular mesh; 3), performing sparse local decomposition on the residual difference of the LA vector and acquiring an LA base; 4), linearly mixing the LA base and acquiring a new LA vector; 5), recovering a connection mapping; 6), estimating the bid price of local orthogonality; 7), restructuring the triangular mesh. The algorithm mainly solves the problem how to perform the sparse local decomposition on a given movement mesh sequence, thereby capturing deformation of a local zone of a movement sequence; moreover, by taking it as the basis, an integrated treatment framework capable of capturing the movement zone and performing gesture edition of the triangular mesh model is structured. The sparse local decomposition and restructuring algorithm is better in locality, applicable to large-scale rotating movement and deformation, overall rotation, and good in robustness.

This invention discloses a method for organizing ring nodes in the broadcast-selection system of peer-to-peer network flow media including: 1, a newly added node P gets an initial member node list from an index server, 2, P utilizes the initial member node to set up its adjacent nodes and distributes them to the corresponding [l d l/alpha] layer of ring based on the distance from the adjacent node to itself so as to constitute an initial logic ring and form far and near hop neighbor nodes, 3, node P transmits position information to a neighbor node on the ring to inform its existence and receives the forward information from the neighbor node to discover new member nodes to ensure there are k nerghbors in the inner ring and 1 neighbor in the outer ring to maintain a stable ring structure and users can quickly find out the neighbor nodes near the target position when dragging, 4, node P stops the current channel and informs all neighbor nodes on the ring about its leaving.

The invention provides a depth image tumbling recognition algorithm based on a wavelet moment. The algorithm comprises the steps of firstly normalizing the distances between all pixel points in an image and the centroid, then conducting coordinate and FFT transformation on the normalized image, finally conducting wavelet transformation on the image to obtain a characteristic vector and recognizinghuman body behaviors in combination with a minimum distance classification method. By testing a large quantity of human body behavior samples, the recognition rate of the algorithm reaches 90%. Through a depth image and translation, scaling and rotation invariance of the wavelet moment characteristics, the privacy problem which exists in traditional video image processing is solved, and the recognition performance is improved. The recognition capability of the human body tumbling behaviors is improved, and the algorithm has better robustness and a very good application prospect.

The invention relates to a nicotinic acid skin reaction image analysis method and equipment. The method comprises the following steps: pasting a nicotinic acid methyl ester patch on the skin of the forearm of a subject; acquiring a nicotinic acid reaction area image of the forearm skin of the subject and analyzing the nicotinic acid reaction area image to generate an identification mark graph of nicotinic acid skin reaction of the subject; obtaining an identification mark graph in the existing diagnosis data as an input to train a prediction model of nicotinic acid skin reaction; inputting anidentification mark graph of the nicotinic acid skin reaction of the subject into the prediction model, and obtaining the prediction result of the nicotinic acid skin reaction of the subject. According to the method, the problem that at present, a rapid and objective quantification tool is lacked for the nicotinic acid skin reaction degree is solved, an image processing and analysis method is introduced, schizophrenia and other severe psychosis serve as research and development service and application objects, and quantitative detection of the nicotinic acid skin reaction degree is achieved.

The invention relates to a lake water volume space-time change analysis method and device based on multi-source satellite data, and the method specifically comprises the steps: S1, obtaining a research region image of a lake through a satellite, preprocessing the research region image, obtaining an image classification map, and obtaining the lake water level through the satellite; and S2, obtaining time series data of the lake water quantity according to the image classification map of the lake and the lake water level, inputting the research area image of the small lake into a KELM model to obtain an image classification map, and inputting the research area image of the large lake into a trained SVM model to obtain an image classification map; the training process comprises the followingsteps: collecting an experimental area image, dividing pixel points on the experimental area image into a water body and a non-water body to form a training sample, training a KELM model and an SVM model by using the training sample, and performing parameter optimization on the KELM model by using a PSO algorithm. Compared with the prior art, the method has the advantages of good stability, good robustness, high efficiency and the like.

The invention discloses a drop height fitting method which comprises the following steps: maintaining an effective test point list; determining whether cubic B-spline fitting is available or not; then performing cubic B-spline fitting or drop test guide; solving the critical drop height; determining whether an effective critical drop height is obtained or not, and then outputting the results or performing drop test guide. According to the invention, conflict points are deleted according to conflict decision rules to obtain effective test points, so that the cubic B-spline fitting can be carried out quickly, and the test efficiency can be improved; meanwhile, a cubic B-spline curve is used for curve fitting for the effective test points, so that fast solution to the critical drop height is realized, and the test efficiency is further improved; the drop test guide is carried out by adopting a successive approximation method, and probable proper test points are prompted for testing, so that the test efficiency is improved.

The invention discloses a vehicle loss detection method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a target image; inputting the target image into a network model, a backbone network of the network model including a Swin Transformer network, the backbone network being used for predicting a damage position coordinate and a damage category of the target image based on the Swin Transformer network; and determining a damage detection result according to the damage position coordinate and the damage category. According to the embodiment of the invention, the Swin Transformer network is used as the backbone network, so that compared with a CNN detection mode, the method is more accurate, and can more effectively positions and identifies a damaged part. The Swin Transformer is used as the backbone network to extract features, so that spatial information relation among pixels of the image and weighted selection of the features can be explored, and better feature extraction and utilization are realized. In addition, the Swin Transformer has the characteristics of locality, translation invariance, residual learning and the like of the CNN, so that the problems of complicated calculation amount and large memory consumption in other visual Transformer schemes can be solved while the performance of the Swin Transformer method exceeds that of the CNN method.

The invention relates to the field of lighting dimming, in particular to a brightness adjusting method and system, a storage medium, a lamp and a lighting system. The method comprises the following steps: correcting an original Gamma curve based on m correction factors to obtain a corrected Gamma curve; and performing Gamma correction on a light source based on the corrected Gamma curve. According to the invention, the original Gamma curve is corrected based on the correction factor with the smooth curve, so that accurate correction and smooth compensation are achieved, and light sensed by human eyes during dimming of the light source subjected to Gamma correction can be linearly changed.

The invention relates to a precise air floating platform, and belongs to the technical field of transportation devices. The air floating platform comprises one or more parallel supporting surfaces, and each supporting surface comprises a positive pressure unit and a negative pressure unit; the positive pressure units comprise pressure outlets, and the pressure outlets provide pressurized gas and are used for generating supporting gas films for supporting a borne object; the negative pressure units comprise pressure inlets, and the pressure inlets provide a pressure value smaller than the environment pressure value and generate adsorption pressure opposite to the supporting gas films; and an atmosphere communicating groove surrounds the periphery of each positive pressure unit and the periphery of each negative pressure unit, and the atmosphere communicating grooves locally discharge the mass flow of the pressure outlets or locally provide the mass flow of the pressure inlets, so that the pressure at the atmosphere communicating grooves is kept as the environment pressure. By adopting the technical scheme, a thin flexible object can be stably, reliably and contactlessly suspended on the air floating platform, and low-friction, high-precision, lossless and efficient supporting, transportation, detection and processing treatment are realized.