71 results about "Scaled correlation" patented technology

The invention relates to a robust target tracking method based on deep learning and multi-scale correlation filtering. The tracking process is divided into a target location part and a scale selection part. In the target location part, the position of a target is located through a convolutional neural network and correlation filtering. In the scale selection part, a scale pyramid is used, and different scales are selected in a matching manner for targets through scale filtering. The multilayer characteristic of the convolutional neural network is taken as a representation model of targets, so the structural and semantic information of targets can be described robustly. Through use of the characteristics of correlation filtering, there is no need to train a classifier online, and the running speed of the algorithm is increased greatly. The idea of scale pyramid is adopted in scale, and correlation filtering matching is performed on targets of different scales to select the optimal scale. The method is of strong robustness to deformation, shading and scale change of targets.

A method for analyzing data, the data characterized by a set of scalars and a set of vectors, to analyze the data into components related by statistical correlations. In preferred embodiments, the invention includes steps or devices for, receiving a set of a scalars and a set of vectors as the inputs; calculating a correlation direction vector associated with the scalar and vector inputs; calculating the inner products of the input vectors with the correlation direction vector; multiplying the inner products onto the correlation direction vector to form a set of scaled correlation direction vectors; and subtracting the scaled correlation direction vectors from the input vectors to find the projections of the input vectors orthogonal to the correlation direction vector. The outputs are the set of scalar inner products and the set of vectors orthogonal to the correlation vector. The steps or devices can be repeated in cascade to form a multi-stage analysis of the data. The invention can also be used with a steering vector preceding the adaptive analysis stages.

The invention relates to an electromyographic signal classification method based on a multi-kernel support vector machine. For a sample with complex distribution, based on the classification performance of a single-kernel support vector machine, the classification accuracy and the quantity of support vectors are easily influenced. The method combines a multi-kernel support vector machine method with a binary tree combination strategy and comprises the following specific steps of: collecting electromyographic signals of the lower limbs of a human body through an electromyographic signal acquisition instrument; denoising the electromyographic signals containing interference noise by using a wavelet coefficient inter-scale correlation denoising method; extracting the features of the denoised electromyographic signals to obtain the features of the electromyographic signals by using denoised wavelet coefficients; and classifying on the basis of the multi-kernel support vector machine. The method can well meet the multi-classification requirement of lower extremity prosthesis control, and takes into account both accuracy and instantaneity, and has broad application prospects in the multi-movement mode recognition of intelligent prosthesis control.

The invention discloses a self-adaptive feature fusion-based multi-scale correlation filtering visual tracking method. The method comprises the following steps: firstly, the correlation filtering is carried out on a target HOG feature and a target color feature respectively by using a context-aware correlation filtering framework; the response values under the two features are normalized; weightsare distributed according to the proportion of the response values and then are subjected to linear weighted fusion, so that a final response graph after fusion is obtained; the final response graph is compared with a pre-defined response threshold value to judge whether the filtering model is updated or not; finally, a scale correlation filter is introduced in the tracking process, so that the scale adaptability of the algorithm is improved. The method can be used for tracking various features. The performance advantages of the features are brought into play, and a model self-adaptive updating method is designed. In addition, a precise scale estimation mechanism is further introduced. According to the invention, the updating quality and the tracking precision of the model can be effectively improved, and the model can be changed in scale. The method is good in robustness under complex scenes such as rapid movement, deformation, shielding and the like.

The invention discloses a short-period power combination prediction method for a variable-weight-coefficient grid-connected photovoltaic power station, and the method comprises the steps: building a plurality of single-prediction models through historical data which is the nearest to a prediction time period, and solving the fitting value of each single-prediction model to each prediction sample point; calculating the weight coefficient of each single-prediction model at each prediction sample point through a gray scale correlation analysis method; carrying out the training and fitting of the fitting value of each prediction sample point and the corresponding weight coefficient through each single-prediction model, and obtaining a BPNN network model; obtaining a single power prediction value of each single-prediction model through the prediction values of the latest meteorological elements in a prediction time period, calculating a time-varying weight coefficient through the BPNN network model, and finally calculating the weighted power prediction value in the prediction time period; carrying out the above steps in a looped manner, and continuously updating the power prediction value in the prediction time period. Compared with the prior art, the method achieves the real-time changing of the weight coefficient of a combination prediction model, and is high in prediction precision.

The invention discloses a time domain high pass non-uniformity correction method based on gray scale correlation, relating to the time domain high pass non-uniformity correction method based on gray scale correlation in the infrared imaging technology field and belonging to the infrared imaging technology field. The time domain high pass non-uniformity correction method based on the gray scale correlation comprises steps of using spatial domain low pass filtering result having edge protection as a correction reference source to perform pre-correction on an inputted image, calculating correction offset value of each frame by combining with time domain high pass filtering, changing the mapping relation between the offset value and the grey scale according to the changing volume of the incident radiation in the same position of each frame, removing the ghost during the correction procedure, and improving the infrared imaging quality. The time domain high pass non-uniformity correction method based on the gray scale correlation can reduce the occurrence probabilities of the ghost and the over-correction of the real-time infrared imaging system non-uniformity correction algorism, improves the infrared imaging quality, reduces the calculation quantity and storage space and facilitates the realization of the hardware.

The invention discloses a device and a device for high-resolution large-viewing-field aerial image forming. The device comprises four cameras with high resolution, a DSP, a displayer and a bracket. A photographing method is realized in a DSP embedded development system and is carried out according to the following steps: firstly, adjacent image overlap proportion is obtained by utilizing system hardware parameters for calculation or calibration, and then direct cutting and splicing are carried out; secondly, a small malposition resulted from the direct splicing is corrected by the typical value removing normalized gray scale correlation calculation method; and finally, according to color difference typical value of splicing regions, differential complement method is adopted for fully correcting the images so as to obtain a high-resolution large-viewing-field aerial image. The invention ensures the large viewing field and realizes the high resolution so that the image forming device can photograph the high-resolution large-viewing-field aerial image at one time.

The invention discloses a CFAR detection method based on gray correlation characteristics in the multi-target environment. According to the method, a background window is adaptively thresholded for clutter truncation, and heterogeneous pixels in the background window are eliminated, and the real clutter is retained to the maximum extent; the maximum likelihood method is utilized to perform two-parameter (the log-mean and the logarithmic standard deviation) estimation on the truncated clutter, and two-dimensional lognormal distribution is utilized to achieve accurate modeling of gray joint probability density between adjacent pixels of the clutter; according to the given false alarm rate, the joint CFAR detection results with different pitches and different directions are obtained, and lastly, the joint CFAR detection results with the different pitches and the different directions are merged to realize CFAR detection based on gray correlation characteristics. The method is advantaged inthat statistical characteristics of the signal-to-noise ratio, gray-scale correlation and the truncated clutter are comprehensively utilized, on the condition that the target detection rate in the multi-target environment is improved, the false alarm rate can be effectively reduced, and the application value is relatively high.

The invention relates to an NSST domain flotation froth image enhancement and denoising method based on a quantum harmony search fuzzy set. The NSST domain flotation froth image enhancement and denoising method comprises the steps: carrying out NSST decomposition on a flotation froth image, and obtaining a low-frequency sub-band image and multi-scale high-frequency sub-bands; performing quantum harmony search fuzzy set enhancement on the low-frequency sub-band image; secondly, for the multi-scale high-frequency sub-bands, removing a noise coefficient by combining an improved BayesShrink thresholding and scale correlation, and enhancing an edge coefficient and a texture coefficient through a nonlinear gain function; and finally, performing NSST reconstruction on coefficients of the processed low-frequency sub-band and each high-frequency sub-band to obtain an enhanced de-noised image. According to the NSST domain flotation froth image enhancement and denoising method, the brightness, the contrast and the definition of the froth image can be improved, and the froth edge is obviously enhanced while noise is effectively inhibited, and more texture details are reserved, and subsequent processing such as froth segmentation and edge detection is facilitated.

The invention discloses a target tracking method based on context correlation and a discriminant correlation filter. The method comprises the following steps: S1, constructing an end-to-end tracking network based on correlation filter, and constructing a tracker for tracking an object with the tracking network as a reference networkS2, generating a feature map by using the first three convolutionlayers of the VGG16 model, and training and learning a context-related filter and a scale-related filter based on the feature map and the context information; 3, training a translation filter in combination with that context-related filt and the feature map, and locating a position of a tracking object with the translation filter; S4, using the scale correlation filter to calculate the proportionof the tracking object based on the position of the tracking object, and locating the position of the tracking object in the next frame in combination with the translation filter, the scale correlation filter, the feature map and the context information; The invention can effectively improve the accuracy and robustness of target tracking.

The invention discloses a m-sequence pseudo random interleaving identification method in a non-cooperated condition. The method comprises the steps of analyzing periodicity in m-sequence pseudo random interleaving, estimating a m-sequence period in pseudo random interleaving, obtaining a mathematical expression of a m-sequence e(t) according to a matrix derivation method, calculating a related function of a pseudo random interleaving signal f(t), calculating a related function of the pseudo random interleaving signal f(t), calculating the power spectrum of intercepted data, calculating a secondary power spectrum, deriving the period in m-sequence pseudo random interleaving, identifying m-sequence pseudo random interleaving in reference to a PN-sequence multi-scale correlation averaging method. The m-sequence pseudo random interleaving identification method can effectively settle a problem of identifying a pseudo random interleaving sequence in the non-cooperated condition. The m-sequence pseudo random interleaving identification method has advantages of simple process, high identification accuracy, high convenience in application, etc. Furthermore communication signal identification capability can be greatly improved.

The invention discloses a multi-scale correlation filtering target tracking method and a multi-scale correlation filtering target tracking device based on response discrimination. The method specifically comprises the steps of determining an initial target position and an initial target size of a current frame image; determining a tracking window size according to the initial target size; extracting a sampling image block in the current frame image based on the initial target position and the tracking window size; performing target detection on the sampling image block, obtaining a response graph corresponding to the sampling image block, and extracting a maximum response peak value of the response graph as a first maximum response peak value; determining a final target position and a final target size in the current frame image according to the first maximum response peak value in response to the fact that the first maximum response peak value is greater than or equal to the first threshold value; according to the method, multi-scale detection does not need to be carried out on each frame of image, so that the calculation amount of target tracking in the video sequence is reduced,the calculation time is shortened, and real-time tracking of the video sequence is realized.

The invention provides a method for carrying out visual tracking through a spatio-temporal context, comprising the following steps of step 1, initializing parameters; 2, training a context-aware filter to obtain a position model; 3, training the maximum scale response value of a scale correlation filter to obtain a scale model; 4, outputting a response diagram by the classifier; discriminating thepeak sidelobe ratio corresponding to the peak value of the response map generated by the correlation filter; 5, comparing the peak-to-peak sidelobe ratio of the response map, and if the peak-to-peaksidelobe ratio of the response map is greater than the peak-to-peak sidelobe ratio, introducing an on-line random fern classifier for re-detection; if the peak value of the response map is less than the peak sidelobe ratio, updating the position model of the step 2 and the scale model of the step 3; if the peak value of the response map is equal to the peak sidelobe ratio, continuing to maintain the current visual tracking state; 6, applying the updated position model and the scale model to the next frame tracking; returning to the step 4.

The invention belongs to the technical field of prediction or optimization, and discloses a prediction method for a power load under complex characteristic influence, and a computer information processing system. The method comprises the steps: inputting historical power load characteristics and influence factor data thereof; carrying out the standardization of an original data sequence, so as toeliminate the interference caused by the difference of units; carrying out the PCA (principal component analysis) of the standardized sequence, and obtaining a PCA expression; calculating the scores of all principal components and an integrated score, obtaining a new data sequence, and carrying out the gray scale correlation analysis of the new data sequence; determining the weight value of each correlation coefficient according to a rule that the weight value is larger for a closer correlation coefficient; carrying out the ordering of the obtained weighted correlation degrees, substituting the weighted correlation degrees into a prediction model, and obtaining a power load characteristic prediction value. The method reduces the analysis complexity of a plurality of influence factors, solves a problem that the information in the influence factors are overlapped, is high in prediction reliability, can meet the requirements of the prediction of the power loads under the complex characteristic influence, so as to solve a problem of accurate planning and scheduling optimization in a power grid in future.

The invention provides an automatic depth correction method based on dual-scale correlation contrast. The method comprises the steps that S1. electric imaging logging data are loaded and the electric imaging logging data are preprocessed; S2. undersampling interpolation is performed on an imaging logging natural gamma curve GR<0><image> according to the sampling interval of a conventional logging natural gamma curve GR<0><log> in conventional logging data; S3. rough correction is performed on an imaging logging natural gamma correction curve GR<1><image> and the conventional logging curve GR<0><log> through first time of contrast under a first scale window; S4. resampling is performed on the conventional logging natural gamma curve GR<0><log> according to the sampling interval of the imaging logging natural gamma curve GR<0><image>; S5. fine correction is performed on the imaging logging natural gamma curve GR<0><image> and the conventional logging natural gamma curve GR<1><log> within depth range of rough correction obtained under the first scale window through the second time of contrast in a second scale window; and S6. an imaging logging natural gamma correction curve GR<2><image> under dual scales is obtained, and then other imaging logging curves are corrected according to the rule of the imaging logging natural gamma correction curve GR<2><image>.

In various embodiments, a prediction application computes a quality score for re-constructed visual content that is derived from visual content. The prediction application generates a frame difference matrix based on two frames included in the re-constructed video content. The prediction application then generates a first entropy matrix based on the frame difference matrix and a first scale. Subsequently, the prediction application computes a first value for a first temporal feature based on the first entropy matrix and a second entropy matrix associated with both the visual content and the first scale. The prediction application computes a quality score for the re-constructed video content based on the first value, a second value for a second temporal feature associated with a second scale, and a machine learning model that is trained using subjective quality scores. The quality score indicates a level of visual quality associated with streamed video content.

The invention relates to a cable porcelain shell terminal infrared image de-noising method considering the inter-scale correlation, which includes the following steps: (1) inputting a cable porcelain shell terminal infrared image to be de-noised; (2) carrying out two-dimensional wavelet decomposition on the infrared image to get a scale coefficient of the highest decomposition layer and wavelet coefficients in three directions of different decomposition layers; (3) calculating the absolute value of the inter-scale correlation coefficient of the wavelet coefficients; (4) using a means clustering method to divide the absolute value of the correlation coefficient obtained in step (3) into a correlation coefficient of effective wavelet coefficients and a correlation coefficient of non-effective wavelet coefficients; (5) directly setting the non-effective wavelet coefficients obtained in step (4) to zero, and retaining the effective wavelet coefficients obtained in step (4); and (6) using the scale coefficient of the highest decomposition layer obtained in step (2) and the processed wavelet coefficients obtained in step (5) to carry out two-dimensional wavelet reconstruction to get a de-noised infrared image. Through the de-noising method of the invention, noise can be removed, and the details of the image can be all retained.

The invention discloses a road extraction method using a non-subsampled contourlet direction field, which mainly overcomes the shortcomings that the road extraction is incomplete and the road is wrongly linked or incompletely linked in the original road extraction method for remote sensing images. The method comprises the following steps: (1) performing non-subsampled contourlet transform (NSCT) on an input remote sensing image to obtain a direct sub-band needing to be processed; (2) estimating the direction field of a direction sub-band and performing directional sub-band correction; (3) performing scale correlation processing on the processed direction sub-band; (4) primarily extracting the road; (5) post-processing the primary road extraction result according to the characteristics of length, direction and shape of road segments; and (6) linking the road segments according to the characteristics of the road segment direction field and the road mutual characteristics. The method can perfectly extract relatively fuzzy road characteristics of the remote sensing images and eliminate interference of the non-road information, ensures relatively accurate extraction result and is applicable to the road extraction application for the remote sensing images.

The invention discloses a self-supervision monocular depth estimation method based on deep learning, and the method comprises the steps: respectively extracting the pyramid features of an original right view Ir and a synthesized left view, carrying out the horizontal correlation operation of the pyramid features to obtain a multi-scale correlation feature Fc, and obtaining a completed multi-scalecorrelation feature Fm; sending the Fm to a visual clue prediction network in a binocular clue prediction module, generating an auxiliary visual clue Dr, reconstructing a right view from the synthesized left view, and optimizing the binocular clue prediction module through image reconstruction loss between the reconstructed right view and a real right view Ir; using a visual clue Dr generated by the binocular clue prediction module for constraining a disparity map D1 predicted by the monocular depth estimation network, and enhancing consistency between the monocular depth estimation network and the disparity map D1 by using consistency loss; and constructing constraints of occlusion guidance to allocate different weights for reconstruction errors of occluded area pixels and non-occluded area pixels.

The invention belongs to the technical field of computer vision, and particularly relates to a target tracking method combining scale adaptation and model updating, which comprises the following stepsof: 1, determining a preliminary search area according to the target state of a current frame; 2, training a scale correlation filter, and estimating the scale change of a target, so as to accuratelyadjust the size of a search region; 3, constructing a training model to obtain a confidence response graph; completing occlusion judgment according to fluctuation of the response diagram; 4, adaptively adjusting the learning rate of the model according to the occlusion judgment; and 5, updating the corresponding training model through a given threshold value. According to the method, the target tracking performance of the typical tracker under the influence of complex factors such as scale change, shielding interference and illumination background is effectively improved.

The invention discloses a physical action attribute analysis and recognition method for intelligent teaching space-time correlation. The method comprises the following steps: tracking and positioningthe position of a learner in each frame according to a real-time visual target tracking mode of multi-scale correlation filtering; And based on a multi-frame image joint segmentation mode, establishing a deformable part model of the learner, and realizing space-time correlation estimation of single-frame overall image action attribute perception. The scheme provided by the invention can improve the spatial-temporal structure prediction and recognition capability of the high-quality physical action education resource visual media of the education resource public service platform, and the structural, continuous and spatial-temporal consistency of physical actions and the like are analyzed to realize the analysis and recognition of the physical actions and the accurate prediction of the change trend of the physical actions.

The invention discloses a crop nutritional deficiency experimental system comprising a water planting barrel, an oxygen-increasing device, a water-supplying device, a nutrient solution-supplying device and a draining device. Crops are water planted in the water planting barrel, ultrapure water is transmitted to the water planting barrel through the water-supplying device, nutrient solution in high concentration is transmitted to the water planting barrel through the nutrient supplying device, the oxygen-increasing device is used for adding oxygen to the nutrient solution in the water planting barrel, the draining device is used for draining waste nutrient solution in the water planting barrel clearly. By the application of the crop nutritional deficiency experimental system, water supplying, nutrient solution supplying, oxygen adding, waste nutrient solution draining are reasonable designed, the efficiency of nutritional deficiency experiment with the water planting device is highly improved, and basic conditions for executing large-scale correlation research are provided. In the process of experiments, parallel experiments are mutually independent and non-interfered, thereby management of experiment is facilitated and the accuracy of experimental data is improved conveniently.

The invention relates to a long-time target tracking method based on a multi-correlation filtering model, and belongs to the field of computer vision. The method comprises the following steps: S1, extracting HOG and HOI features of a video image, and training a long-term correlation filter; s2, during a tracking process, judging whether target tracking fails or not by utilizing a maximum responsevalue generated by a long-time correlation filter and a target and a detection threshold value; if the target tracking succeeds, estimating the translation of the target by adopting an optimal displacement correlation filter in an MCCT algorithm and obtaining the position information of the target, and if the target tracking fails, activating an online detector to reposition the target and takingthe detection result of an online classifier SVM as the position information of the target; s3, after determining the translation position of the target, determining the scale of the target in the frame by using a scale correlation filter; and S4, finally, updating the filter model under the condition of meeting the target updating condition. According to the invention, the time overhead is reduced, and the performance is superior.

The invention discloses an anti-shielding self-adaptive target tracking method and system, a scale correlation filter is trained while a position correlation filter is trained, scale self-adaptive transformation can be realized, if the transformation does not exist, the size of a target frame is not changed in the training process and is the same as the size of a rectangular frame which is determined manually at the beginning, and the target frame is not changed in the training process. However, after scale transformation, the size of the target frame can be automatically changed along with the distance of the target, the target frame becomes smaller when the target moves farther from the camera, and the target frame becomes larger when the target moves closer, so that the accuracy and robustness of the whole algorithm are improved. According to the target tracking method, an adaptive model updating strategy is adopted, and whether the target is shielded or lost or not is detected by calculating a PSR value, so that a search area is expanded, the problem that tracking cannot be continued once the target is shielded or lost due to target movement and the like in a traditional target tracking method is solved, and the target tracking efficiency is improved. And the continuity and reliability of target tracking are improved.

The invention discloses a forest fire smoke detection method based on wavelet change image enhancement and multiple characteristics, and belongs to the field of fire smoke detection methods. An existing forest fire smoke detection method has the problem of missing report or false report. The invention discloses a forest fire smoke detection method based on wavelet change image enhancement and multiple characteristics. The method comprises: carrying out wavelet decomposition on the smoke image; gray scale correlation degree calculation and wavelet coefficient correction are carried out on the wavelet coefficient; smoke image noise under a complex background is suppressed; Enhancing smoke images, Fractal features of a smoke image and image texture features based on a gray level co-occurrencematrix are calculated, the difference between the fractal features and the texture features of an image smoke area and a non-smoke area is utilized, the features are input into an SVM for training, and effective detection of a forest fire smoke area is achieved through machine learning and a large sample size. The fire smoke video detection method is stable in performance, low in missing report rate and high in accuracy.