44 results about "Local regression" patented technology

The invention discloses a TRMM (Tropical Rainfall Measuring Mission) satellite rainfall product downscaling method based on M5-Local and a multi-environmental factor variable. An M5-Local regression thought is adopted, and a plurality of environment stress factors with high spatial resolutions are fully utilized to improve the spatial resolution of the product. Firstly, 1km environmental variable factors, such as eight pieces of data including a vegetation index, a digital elevation model, a daytime surface temperature, a night surface temperature, a terrain moisture index, a gradient, a slope aspect and a slope length gradient, are subjected to aggregate calculating to obtain 25km which serves as an independent variable, corresponding 25km-resolution TRMM data is used as a dependent variable, an M5-Local method is adopted for modeling, the intercept of a regression modeling equation of 1km spatial resolution and a slope parameter corresponding to each environment factor variable are predicted, and a 1kmTRMM rainfall value is calculated. A downscaling result based on an M5-Local model is obviously superior to a downscaling result of a conventional regression model.

The invention relates to a color constancy method based on a texture pyramid and regularized local regression, belonging to the technical field of image processing. The color constancy method based on a texture pyramid and regularized local regression comprises the following steps: establishing an algorithm candidate set for the fusion algorithm according to different parameter selections througha Minkowski paradigm based on common color constancy, segmenting an image into overlapped blocks according to different resolution ratios, describing the texture features of each block by adopting Weibull distribution parameters, selecting K images of which the texture features are similar to the texture features of the given images to be measured from a training database according to the definedimage similarity norm, estimating the illumination of the images to be measured by adopting the regularized local regression according to the K reference images in the fusion stage, and finally mapping the images to be measured to the standard white light by using a Von Kries model according to the illumination estimation value to obtain the corrected images. The invention effectively improves the illumination estimation accuracy and can be widely applied to target tracking, object recognition and image retrieval, and the like.

The invention relates to a sports video tactical behavior recognition method based on a space-time local mode, and belongs to the technical field of video content understanding. A local space-time mode is used for realizing the tactical video behavior recognition of teamwork match in sports videos. The sports video tactical behavior recognition method is different from the existing method using track features for tactical behavior recognition. For video frame sequence images, a measure of improving a space-time local regression kernel to be used as a feature detector for detecting a movement significance region is provided; and the movement significance region is directly used as a feature word for building a visual feature word bag model and is used for tactical behavior recognition. The defect of influence on the multi-target track extraction due to factors such as complicated backgrounds in the multi-target track extraction process is overcome; the local space-time mode features and the space-time distribution of the local space-time mode features are used as tactical expression; the complexity of the recognition method is reduced; and meanwhile, the practicability of the method is improved. By aiming at large data volume of the videos, the detecting efficiency of a detecting operator is effectively improved.

The invention discloses a three-dimensional human body posture estimation method based on feature fusion and sample enhancement, and relates to a three-dimensional human body posture estimation and performance optimization method. The method comprises the following steps: firstly, carrying out body part classification and pixel point three-dimensional coordinate regression on a human body in a picture by adopting a full convolutional network based on a candidate region; secondly, auxiliary network sample enhancement is adopted, and signal supplementation is conducted on sample positions without initial annotations; and finally, performing feature fusion on the model and an existing 2D attitude recognition model with a good effect, and performing advantage complementarity with local regression coordinates from the perspective of global attitude. A human body posture estimation framework based on multi-task parallelism is constructed through a feature fusion technology, and an effectivetheory and method are provided for advantage complementation of two-dimensional posture recognition and three-dimensional posture recognition; an auxiliary network based on data enhancement is established in a mode of simulating semi-supervised learning, and a new thought is provided for improving the generalization ability of a posture recognition model.

This disclosure provides methods, apparatus and systems for performing image processing regression for approximating multidimensional color transformation. According to an exemplary method, a shaping matrix is selected to minimize a cost function associated with a local linear regression representation of the color transformation. In addition, an alternating least squares algorithm is utilized to jointly optimize regression and shaping parameters.

The invention belongs to the field of image processing, and aims to establish an efficient non-reference stereo image quality evaluation method based on local-to-global feature regression. The stereoimage quality evaluation method is more accurate in quality prediction, does not need to depend on an original reference image, can replace a subjective evaluation result to a certain extent, improvesthe efficiency of stereo image quality evaluation work, and provides certain convenience for subsequent work. Therefore, the technical scheme adopted by the invention is as follows: the non-referencestereo image quality evaluation method based on local-to-global feature regression comprises the following steps: firstly, respectively endowing image blocks of a left viewpoint and a right viewpointwith different tags through a feature similarity FSIM algorithm, guiding a local regression neural network comprising a left channel and a right channel to perform pre-training by utilizing the calculated tags, and storing parameters of the trained local regression neural network. The method is mainly applied to image processing occasions.

The present invention discloses an identification method of an oil depot target in a remote sensing image. The method comprises the steps of firstly calculating the phase spectrum significance of a whole scene, and extracting all interested areas possibly containing the target in the scene according to the phase spectrum significance; in the feature extraction, adopting a local regression nuclear model to calculate the local structural features of the interested areas point by point, and generating a feature descriptor capable of describing the structure of the target; at a target detection stage, carrying out the similarity measurement by the cosine similarity, calculating the similarities of the interested areas and an oil depot sample image, utilizing the positive and negative sample distinguishing ability of the feature descriptor and the features of the similarity surfaces to construct a decision network having the adaptive capability, obtaining the preliminary results of the target detection by the decision network, and removing the redundant preliminary results by a non-maxima suppression algorithm, thereby obtaining a final target detection result. A universal detection method of the oil depot target in the remote sensing image provided by the present invention is good in multi-scale and multi-view angle target identification effect.

The invention discloses an image super-resolution reconstruction method based on a local regression model. The method comprises the following steps: at first, carrying out Gaussian low pass filtering on an input low resolution image to obtain a low frequency band image thereof, carrying out bicubic interpolation to obtain an approximate low frequency band image of a high resolution image; then, applying a one-order regression model to each image block in the low frequency band image of the high resolution image during reconstruction, wherein a mapping function between high / low images in the regression model can be obtained by a machine learning method of an input image, namely, sampling corresponding positions of the input low resolution image and the low frequency band image thereof to obtain sampling image blocks of corresponding positions, and carrying out dictionary training; and finally, respectively applying the one-order regression model to non-local self-similar blocks of the reconstructed image blocks, and carrying out weighted integration to obtain reconstructed high resolution image blocks. By adopting the method provided by the invention, no external image model is required, a prior model is obtained by learning the input image, and the high resolution image reconstructed by the model has better subjective and objective reconstruction effects.

The disclosure provides an approach for denoising (also referred to as “filtering”) rendered images. In one embodiment, a denoising application takes as input rendered images and feature buffers that encode image information such as surface positions, surface depths, surface normals, surface albedos, and distances to the camera. For each pixel in a received image, the denoising application performs a first-order regression in a predefined neighborhood of the pixel to find a linear combination of pixel features that fits pixel colors in the predefined neighborhood. In such a first-order regression, the local regression weight of each pixel in the neighborhood may be determined using a metric which computes distances based on color values in patches around pixels being compared. In another embodiment, collaborative filtering may be performed in which filtered output from the first-order regression in each neighborhood is averaged with filtered output from overlapping neighborhoods to obtain a final output.

The present invention provides a kind of multivariate time series forecasting method, this method comprises: Step 1, based on local regression seasonality and trend decomposition algorithm, decompose multivariate time series seasonally into trend sequence, cyclic sequence and irregular sequence; Step 2, Use linear or nonlinear regression algorithm to predict the trend sequence to obtain the predicted value of the trend sequence; step 3, based on the combination model, predict the cyclic sequence that introduces external factors and historical variables respectively to obtain the initial predicted value of the cyclic sequence, using the feedforward neural network The initial predicted value is fused to obtain the predicted value of the cyclic sequence; step 4, the predicted value of the trend sequence and the predicted value of the cyclic sequence are summed to obtain the predicted value of the multivariate time series. The invention solves the problem of prediction accuracy of multivariate time series, can more reasonably analyze the characteristics of multivariate time series and the influence of external factors, and obtain more accurate time series prediction results.

The invention relates to a modeling method for soil heavy mental content based on a forward interval partial least squares algorithm. The method comprises the following steps: dividing the whole heavymetal fluorescence spectrum area into k wavebands in equal width; establishing a partial least squares model for each sub-band, thereby acquiring k local regression models; measuring the model precision on the basis of RMSECV value; taking the waveband corresponding to the local model in the highest precision as a first selected waveband and taking the local model as a first sub-model; jointing the other (k-1) wavebands with the first selected waveband in turn, thereby acquiring (k-1) local models; selecting the waveband corresponding to the local model with the lowest RMSECV value as a second selected waveband and taking the local model as a second sub-model; repeating the processes till jointing wall the wavebands; inspecting the RMSECV value of each sub-model and selecting the sub-model with the optimal performance from all the sub-models: taking the section combination corresponding to the minimal RMSECV as the optimal combination. According to the invention, the spectrum range used by the model is optimized and the precision of the soil heavy mental quantitative detection model is increased.

This disclosure provides methods, apparatus and systems for performing image processing regression for approximating multidimensional color transformation. According to an exemplary method, a shaping matrix is selected to minimize a cost function associated with a local linear regression representation of the color transformation. In addition, an alternating least squares algorithm is utilized to jointly optimize regression and shaping parameters.

The invention belongs to the field of image processing and aims to establish an efficient no-reference stereo image quality evaluation method. The quality prediction is more accurate and the efficiency of three-dimensional image quality evaluation work is improved. According to the non-reference stereo image quality evaluation method based on local to global feature regression, firstly, differentlabels are given to image blocks of a left viewpoint and a right viewpoint through a feature similarity FSIM algorithm, the calculated labels are used for guiding networks of a left channel and a right channel to conduct pre-training at the same time, and therefore local regression of features is achieved; then, a fusion channel is added on the basis of the left channel and the right channel to form a global regression network, on the basis of the pre-training model, a subjective evaluation value DMOS serves as a label to guide network training, network parameters are finely adjusted, and therefore global regression of features is achieved; and the quality of the three-dimensional image is subjected to feature extraction and prediction by the trained global regression network. The method is mainly applied to design and manufacturing occasions.

The invention relates to a local scale landslide susceptibility prediction method based on hierarchical Bayesian. The method mainly comprises four steps of extracting a basic evaluation unit; calculating the regional scale weight of the evaluation factor; establishing a spatial local logistic regression model, fitting spatial heterogeneity of the evaluation factor in a local scale, realizing information fusion of two scales under a hierarchical Bayesian framework, and verifying the model; and carrying out local-scale landslide susceptibility evaluation by utilizing the spatial local logistic regression model. According to the invention, the geographical detector is used to calculate the overall trend of each evaluation factor in the regional scale; fitting spatial heterogeneity of the evaluation factor contribution rate in a local scale by applying a spatial local regression model; and based on a hierarchical Bayesian framework, information fusion is carried out on the overall trend and spatial heterogeneity of the evaluation factor, so that the problem that the stability of the overall trend and heterogeneity of the local scale of the evaluation factor cannot be considered at thelocal scale is solved, and regional local scale landslide susceptibility evaluation is realized.

This disclosure provides methods, apparatus and systems for performing image processing regression for approximating multidimensional color transformation. According to an exemplary method, a shaping matrix is selected to minimize a cost function associated with a local linear regression representation of the color transformation.

The invention discloses a sparse coding method based on local and global regularization, pertaining to the technical field of sparse coding (SC). The method utilizes local regression regularization tofind out a potential geometric structure of data. Concretely speaking, the overall data space is divided into a large number of partial regions. Each sample is linearly represented by a sample of a partial region to which each sample belongs, which is called as a local learning conception. On the basis of the local learning conception, a kernel global regression method is utilized for grabbing aglobal geometric structure of data. Therefore, partial and global regularization are utilized for obtaining an inherent geometric structure of data.

The invention discloses an adaptive rendering method based on weight local regression. The method comprises the following specific steps: filtering and reconstructing an image by use of local regression; constructing a simplified feature space by use of truncated singular value decomposition (TSVD) so as to eliminate noise carried by a feature vector; respectively predicating an optimal feature bandwidth bj and a shared bandwidth h by use of a two-step bandwidth selection algorithm; and distributing available light samples to areas with high errors by use of an iteration method based on a local simplified feature subspace k. The method provided by the invention is a novel adaptive rendering technology and can be applied to processing rendering effects of multiple types in an efficient and robust manner. Compared to prior similar arts, the method has the following advantages: the time for generating results with the same quality is obviously reduced, and the result image effect generated based on equal calculation time and calculation cost is far better than that of the similar arts.

The invention discloses an object identification method based on a context information propagation local regression kernel, comprising steps of obtaining local structure information in a given template image by utilizing a local regression kernel, giving consideration to the context information of the local area, searching a characteristic area similar to a template image in the object image through the correlation between context information and the local information, based on the above arrangement, utilizing a local reservation projection to perform dimensionality reduction on image characteristics in order to reduce the redundancy of the characteristic data quantity, adopting similarity among correlated characters of cosine matrix similarity measurement to find similar character areas, and, at last, utilizing a non-maximum value suppression method to suppress unremarkable area characteristics to obtain a final recognition. The object identification method based on a context information propagation local regression kernel improves accuracy and efficiency of the target identification.

The invention discloses a dense correspondence prediction method based on non-rigid point clouds. The method comprises the following steps: respectively extracting geometric features of a three-dimensional template and point clouds by using a graph convolutional neural network and a plurality of set abstraction layers; utilizing a global regression network to deduce global displacement according to the associated global features of the template and the point cloud; utilizing a local feature embedding technology, introducing an attention mechanism, and fusing local depth features of the point cloud and geometric features of the graph; utilizing a local regression network to predict displacement increment; utilizing a weak supervision fine tuning method to process the real point cloud, and unifying the real point cloud and the two-stage regression network in a complete framework. The local geometric features of the point cloud are fully utilized, the attention strategy is adopted to improve the corresponding precision, the weak supervision fine tuning method is adopted to robustly process the real point cloud, and the conditions that the prediction model is unreasonably distorted and is obviously inconsistent with the input shape due to the lack of training data are effectively improved.

The invention discloses a human eye fixation point extracting method combining global colors and local structure features. The human eye fixation point extracting method comprises the following steps:reading an image to be detected; uniformly quantizing pixel colors in an image in an RGB color space in order to extract global color features; calculating a global saliency map in the entire image range based on a color contrast degree; then, extracting the brightness information of the image; calculating a local regression kernel of each pixel in a brightness channel in order to extract the local structure features; calculating a local saliency map in a local range based on a structural contrast degree; and finally, fusing the global color saliency map and the local structure saliency map to obtain an initial human eye significance degree graph, adjusting the initial human eye significance degree graph in conjunction with center prior, and performing Gaussian filter to obtain a final human eye fixation point position. In conjunction with the image saliency map obtained by the global color features and the local structure features, the significance information in the image is expressed better, and the accuracy and robustness of human eye fixation point extraction are improved.

The invention discloses a method for measuring the pose of a cylindrical object with a circular arc surface. The method can be used for measuring the spatial position coordinates and the spatial attitude angle of a cylindrical object having a circular arc surface of arbitrary size. According to the method, ellipse fitting is carried out on measurement data point groups sequentially, so that data points in contour data of any complex section which belong to contour data point of a cylindrical object with a specific radius are identified; ellipse fitting is carried out on the middle data points of an identified elliptic arc, and a distance between data points at two sides and an ellipse obtained through fitting is calculated, and measurement data which do not belong to the elliptic arc can be accurately eliminated; elliptic arc data points are processed through using a local regression method and a cubic spline interpolation method, so that the data can be smoothened and homogenized fast and effectively; ellipse fitting is carried out on the processed data points, so that five characteristic parameters of the ellipsis shape of the section of the cylindrical object can be obtained accurately; and spatial calculation processing is performed on the parameters, so that the pose information of the cylindrical object can be obtained finally.

The invention discloses a regression method for a load rate-efficiency curve of a boiler. The method comprises the following steps: S1, obtaining a load rate-efficiency data set according to operationdata of the boiler; s2, performing weighted polynomial fitting on data points in the load rate-energy efficiency data set according to a local regression algorithm to obtain corresponding regressiondata points; and S3, forming a regression curve of the boiler load rate-efficiency curve through all the regression data points. The invention further discloses a regression device for the boiler loadrate-efficiency curve. The device comprises a data acquisition module, a regression fitting module and a curve forming module. According to the method, the load rate-energy efficiency curve of the gas-fired boiler is regressed by utilizing a local weighted regression algorithm, so that the influence of abnormal data on a fitting curve can be well reduced, and a more accurate regression curve is obtained. Therefore, a better fitting effect is obtained in a practical problem processing process.

The invention relates to a color constancy method based on a texture pyramid and regularized local regression, belonging to the technical field of image processing. The color constancy method based on a texture pyramid and regularized local regression comprises the following steps: establishing an algorithm candidate set for the fusion algorithm according to different parameter selections through a Minkowski paradigm based on common color constancy, segmenting an image into overlapped blocks according to different resolution ratios, describing the texture features of each block by adopting Weibull distribution parameters, selecting K images of which the texture features are similar to the texture features of the given images to be measured from a training database according to the defined image similarity norm, estimating the illumination of the images to be measured by adopting the regularized local regression according to the K reference images in the fusion stage, and finally mapping the images to be measured to the standard white light by using a Von Kries model according to the illumination estimation value to obtain the corrected images. The invention effectively improves the illumination estimation accuracy and can be widely applied to target tracking, object recognition and image retrieval, and the like.

The invention provides a hemodialysis dry weight intelligent evaluation device. The device is used for evaluating the dry weight required by a patient to be subjected to hemodialysis, and is characterized by comprising: a body weight overall trend model building part which is used for building a linear model corresponding to the relationship between a hemodialysis time point and a body weight after differential hemodialysis, and solving the optimal parameters of the linear model based on a least square method so as to form a body weight overall trend model of the individual patient to be hemodialysis; a local regression model building part which is used for acquiring a data set to be weighted based on a Gaussian kernel function method and carrying out local regression correction on the body weight overall trend model so as to form a weight local regression model; a dry weight prediction part which inputs the current hemodialysis time point into the body weight overall trend model and the local regression model, and acquires predicted dry weight in combination with the body weight index average after the current hemodialysis; and a dry weight adjusting part which is used for adjusting the predicted dry weight according to correction factors corresponding to the characteristics of the patient so as to obtain the dry weight of the patient to be subjected to hemodialysis.

The disclosure provides an approach for denoising (also referred to as “filtering”) rendered images. In one embodiment, a denoising application takes as input rendered images and feature buffers that encode image information such as surface positions, surface depths, surface normals, surface albedos, and distances to the camera. For each pixel in a received image, the denoising application performs a first-order regression in a predefined neighborhood of the pixel to find a linear combination of pixel features that fits pixel colors in the predefined neighborhood. In such a first-order regression, the local regression weight of each pixel in the neighborhood may be determined using a metric which computes distances based on color values in patches around pixels being compared. In another embodiment, collaborative filtering may be performed in which filtered output from the first-order regression in each neighborhood is averaged with filtered output from overlapping neighborhoods to obtain a final output.

The invention relates to a domain generalization image classification method based on low-rank constraint local regression, and the method comprises the steps: (1), obtaining a source domain, and decomposing the source domain into a plurality of sub-domains; (2) training according to each sub-domain to obtain a classifier; step (3), constructing a low-rank constraint local regression framework according to the classifier obtained by training each sub-domain; and step (4), classifying the image based on the low-rank constraint local regression framework. The method can effectively classify the images, and can be applied to domain self-adaption.

The invention discloses a single image super-resolution reconstruction method based on TV priori. The method comprises the following steps: firstly, carrying out up-sampling preprocessing on an imageto be improved in resolution by utilizing a bicubic interpolation algorithm omoms3; extracting TV prior information from the preprocessed image according to 28 TV direction templates; and finally, introducing TV prior information into a non-local regression framework, so that the texture and edge information of the super-resolution image can be better reserved, and the phenomenon of insufficient image edge information generated by a traditional interpolation algorithm can be overcome, and therefore, a super-resolution reconstructed image can be obtained, and a super-resolution effect can be greatly improved.