150 results about "Bayesian framework" patented technology

Visual tracking of multiple objects in a crowded scene is critical for many applications include surveillance, video conference and human computer interaction. Complex interactions between objects result in partial or significant occlusions, making tracking a highly challenging problem. Presented is a novel efficient approach to tracking a varying number of objects through occlusion. The object tracking during occlusion is posed as a track-based segmentation problem in the joint-object space. Appearance models are used to interpret the foreground into multiple layer probabilistic masks in a Bayesian framework. The search for optimal segmentation solution is achieved by a greedy searching algorithm and integral image for real-time computing. Promising results on several challenging video surveillance sequences have been demonstrated.

Systems and Methods for multi-modal or multimedia image retrieval are provided. Automatic image annotation is achieved based on a probabilistic semantic model in which visual features and textual words are connected via a hidden layer comprising the semantic concepts to be discovered, to explicitly exploit the synergy between the two modalities. The association of visual features and textual words is determined in a Bayesian framework to provide confidence of the association. A hidden concept layer which connects the visual feature(s) and the words is discovered by fitting a generative model to the training image and annotation words. An Expectation-Maximization (EM) based iterative learning procedure determines the conditional probabilities of the visual features and the textual words given a hidden concept class. Based on the discovered hidden concept layer and the corresponding conditional probabilities, the image annotation and the text-to-image retrieval are performed using the Bayesian framework.

The invention discloses a moving object detecting method based on a Bayesian frame and an LBP (Local Binary Pattern), which relates to the technical field of intellectualized video monitoring. The method comprises the following steps of: 1. extracting a video frame in a video stream; 2. preprocessing the video frame and eliminating the interference of fine light ray change and other disturbance; 3. carrying out moving object detection based on a background model and the Bayesian frame; before detecting, filtering a pixel point with unchanged time difference and background difference and degrading color dimensionality in modeling, wherein the speed of the whole method is improved, and the background model comprises a color characteristic and a symbiotic color characteristic and can favorably detect a moving and static object in a video; and 4. removing a shadow, detecting the texture information of the obtained moving object area and the texture information of a background image corresponding area by utilizing an LBP descriptor and comparing the differences of the texture information for removing the shadow in moving object detection.

A computer implemented method for joint image registration and reconstruction of a plurality of images includes providing the plurality of images, modeling the plurality of images, including reconstructing bone and soft-tissue in respective images of the plurality of images, performing a hierarchical free-form registration of models of the plurality of images to determine a jointly registered and reconstructed image with successive accuracy adjustment according to a registration error, and outputting the registered and reconstructed image.

An improved system and method for matching objects belonging to hierarchies is provided and an optimal matching between two feature spaces organized as taxonomies may be learned. The matching may be performed through a multi-level exploration of the hierarchical feature spaces by using multi-armed bandits where the arms of the bandit may be dependent due to the structure induced by the taxonomies. Upon the arrival of an object assigned to the first taxonomy, multi-armed bandits may be run at multiple levels of the taxonomies to select an object assigned to the second taxonomy. Then shrinkage estimation may be performed in a Bayesian framework to exploit dependencies among the arms by estimating payoff probabilities from a beta-binomial model to update payoff probabilities for matching objects from the taxonomies.

The invention discloses a continuous and stable tracking method of a weak moving target in a dynamic background. The method includes acquiring the video data, and processing each frame of image in the following steps of acquiring the position coordinate of the moving target to be tracked in the current frame of image and determining the target tracking frame according to the position; establishing a spatial context model of the current frame of image for the area in the target tracking frame with a Bayesian framework; performing convolution calculation by means of the spatial context model of the current frame of image and the next frame of image to obtain a confidence map of the position of the moving target to be tracked in the next frame of image, the position with the greatest confidence is the position of the moving target to be tracked in the next frame of image; based on the double threshold moving target crisis determination, determining that when the moving target to be tracked is not shielded or lost, outputting the moving target position in the next frame of image when the tracking of the current frame of image is completed; otherwise, updating the target tracking frame and re-checking. The method realizes the continuous and stable tracking of the target under the condition of background interference and shielding.

A Bayesian formulation for coupled surface evolutions in level set methods and application to the segmentation of the prostate and the bladder in CT images are disclosed. A Bayesian framework imposing a shape constraint on the prostate is also disclosed, while coupling its shape extraction with that of the bladder. Constraining the segmentation process improves the extraction of both organs' shapes.

The invention relates to an image segmentation method having multi-layer segmentation networks based on Bayes framework with edge prior information. The method herein includes the following steps: S1.performing general classification on input data by using a full convolutional neural network, outputting scoring graphs of all types that have the same sizes of the input data, also extracting an implicit edge image from an internal feature layer of the full convolutional neural network; S2. extracting an explicit edge image from the input data by using an edge detection network; S3.performing first restraining on the types that are obtained from S2 by using domain transformation and conditional random field, obtaining an initial segmentation image; S4. transforming the explicit edge image obtained from S2 to an edge distance image; and S5. inputting the edge distance image to defined domain transformation, performing second edge restraining on the initial segmentation image that is obtained from S3, and obtaining a final segmentation result. According to the invention, the method can extract edge prior information through an external edge network and performs edge region segmentationand filtering on the result from the general segmentation by using the edge prior information, so that the method herein can increase the accuracy in segmenting a SAR image.

The invention provides a data association method for utilizing a sensing network to carry out multi-target tracking. The method comprises the steps of: clustering sensor nodes; collecting measurement information of targets by the nodes; calculating state discrete time equations of the targets; calculating state prediction results of the targets; calculating the posterior probability of a target k in a node I; carrying out target tracking by means of a Bayes framework; calculating the posterior probability distribution of the targets; and repeating the above steps to finish target tracking. In order to carry out accurate tracking on multiple targets, the invention provides the new method to solve the data association problem in wireless sensor network multi-target tracking, the method is based on the Bayes framework, the target position state information is utilized, and the speed state information is also utilized; and compared with the modeling methods, the method provided by the invention is low in calculation amount, high in precision and is capable of accurately carrying out multi-target tracking under a low-cost condition especially in the aspect of cross and maneuvering target tracking.

The invention relates to a remote sensing image airport detection method based on a weak supervised learning frame. The method comprises the steps that firstly, the saliency of image blocks in a positive sample, the similarity of image blocks in a positive sample set and the inter-class difference between the positive sample and a negative sample are obtained, then, the Bayesian frame is utilized for fusing the three classes of information to obtain an initial positive and negative training set, then, iteration training of the training set is utilized for obtaining a final stable airport detector, the airport detector is used for detecting an airport of a tested image, and finally the airport detection result with better accuracy and robustness is obtained.

Systems and methods for shape registration are described. In one aspect, training shape vectors are generated from images in an image database. The training shape vectors identify landmark points associated with one or more object types. A distribution of shape in the training shape vectors is represented as a prior of tangent shape in tangent shape space. The prior of tangent shape is then incorporated into a unified Bayesian framework for shape registration.

The invention provides a video tracking method based on local background learning. The video tracking method includes the steps that the time-space relationship between a target to be tracked and the local background of the target is modeled through the Bayes frame, a plurality of multi-dimensional images of the target are simultaneously collected through the time-space relationship between the modeled target and the local background, and dimensions of the collected multi-dimensional images of the target are reduced through a random sensing matrix meeting compressed sensing conditions to obtain feature vectors of the multiple multi-dimensional images; according to the feature vectors of the multiple multi-dimensional images, the multi-dimensional images with the dimensions reduced are classified through a naive Bayes classifier, and the position where the target appears is estimated according to a likelihood confidence image of the target position; based on target structure constraint conditions, a collector outputs the target with the maximum degree of overlapping with the previous frame target tracked successfully as the final tracking target. The video tracking method is suitable for video target tracking under complex conditions, and is high in discernment capacity and tracking accuracy.

The invention discloses a knowledge base completion method based on variational interference and a tensor neural network. Tensor decomposition thought and a Bayesian framework are introduced into the knowledge base completion method, and the prior knowledge of implicit variables is considered. The interaction between the implicit variables is discussed and nonlinear expression is performed through the neural network, and consideration of nondeterminacy is added. The accuracy of the knowledge base method is improved obviously, and compared with the existing technology, the knowledge base completion method has been greatly improved.

The invention relates to a hand area segmentation method deeply integrating significance detection and prior knowledge. This method combines the significant pattern of a hand area at pixel level with the significant pattern of a hand area at the regional level to make the detection algorithm for the entire hand area achieve higher robustness and accuracy. The method comprises the following steps: using an introduced Bayesian framework to obtain the confidence degree of each pixel in the hand area; combining the relevant techniques such as threshold segmentation, and finally obtaining a hand area segmentation result with high accuracy. The invention overcomes the shortcomings of a traditional hand area method which can only be applied to relatively simple background and non-near-skin color interference scene, and the method can obtain a hand area segmentation image accurately even in the case of various disturbances such as non-uniform illumination, near skin color background and face noise, therefore, making it applied with great prospects.

The invention relates to a multi-maneuvering target tracking method based on iteration adaptation. According to t he invention, based on a data association problem and a target maneuvering problem in multi-maneuvering target tracking processes, a confidence propagation algorithm and a jump Markov system fixed interval smoother are used; under the unified framework of the variational Bayesian, united optimization is performed on target states and association matrixes through an iteration mechanism, so tracking precision of the maneuvering target is improved; and compared with the current multi-maneuvering target tracking algorithm, the provided method is characterized in that based on the iteration feedback idea, the data association problem and the target maneuvering problem are processed in an united manner under the unified variational Bayesian framework so the method has certain theoretical and actual engineering meanings in the target tracking field.

The invention discloses a seismic random inversion method and device based on multi-point geostatistical prior information. The method comprises the following steps of determining a to-be-inverted profile, well data and a training image according to a known three-dimensional geological model; determining a rock phase probability distribution model of the to-be-inverted profile according to the well data and the training image; under the constraint of the rock phase probability distribution model, determining a prior probability density function of elastic parameters corresponding to a to-be-inverted seismic trace on the to-be-inverted profile according to the well data and a probability density function and a variation function of the elastic parameters corresponding to different rock phases obtained by performing petrophysical statistics on the well data; and in the Bayesian framework, determining an inversion result according to the prior probability density function of the elastic parameters obtained under the constraint of the rocky phase probability distribution model and seismic data. The seismic random inversion method and device provide more accurate prior information for inversion, are beneficial to improve the accuracy and resolution of the inversion result, and make the inversion result more in line with actual production requirements.

The invention discloses a compressed sensing image reconstruction method based on relevance vector grouping, which mainly solves the problems of inaccuracy and low robustness of compressed sensing image reconstruction. The realization process is as follows: 1) receiving an observation matrix and an observation vector; 2) obtaining an initial relevance vector by the observation vector and a sending matrix; 3) dividing the relevance vector into sub-relevance vectors according to the spatial neighbourhood relationship of wavelet coefficients; 4) adding a component in each sub-relevance vector and sequencing the components; 5) updating the reconstructed wavelet high-frequency coefficients and observation vectors on the basis of a Bayesian framework according to the sequencing order; 6) carrying out invert wavelet transform on the reserved low-frequency wavelet decomposition coefficients and the reconstructed high-frequency wavelet coefficients to obtain a reconstructed image. Compared with OMP and BEPA methods, the compressed sensing image reconstruction method based on relevance vector grouping disclosed by the invention has the advantages of high quality and good robustness of the reconstructed image, and can be used for reconstruction for natural images and medical images.

The present invention provides a block sparse compressive sensing based infrared image reconstruction method the method comprises : the step of block sparse representation, the continuous infrared small target image sequence is shown having a block sparse characteristics of the signals; the time correlation modeling step in bayesian framework infrared small target for continuous the data between the image sequence for modeling time correlation; step of outputting a signal reconstruction, sparse learning algorithm is adopted to realize the reconstruction of the infrared small target image, and a final estimated value of the output signal of the present invention also provides a block sparse compressive sensing based infrared image reconstruction system. The technical scheme provided by the invention in the compressed sensing reconstruction process, utilizing a continuous signal in time correlation signals are reconstructed, the single frame signals can be improved, and the reconstruction accuracy significantly reduces the run time of the algorithm.

The invention relates to a variational Bayesian self-adaptive filtering method. For the target status estimation problem in a case of unknown process noise time-varying, inverse Wishart distribution is used for a priori modeling on process noise covariance through introducing a new hidden variable, and target status, the process noise covariance and the hidden variable are jointly optimized undera variational Bayesian framework through an iterative mechanism to improve tracking accuracy of the target status. Compared with the existing filtering algorithms, the method of the invention directlyselects the conjugate a priori distribution of the process noise covariance, carries out iterative optimization of a target status estimation result under the variational Bayesian framework, obtainsa new self-adaptive filtering algorithm, and has certain theoretical and practical engineering significance for the field of filters.

The invention discloses a salient image extraction processing method and system. The method comprises the steps: analyzing the characteristics of the current image through an RGB channel so as to obtain the RGB salient characteristics of the current image; analyzing the characteristics of the current image through the Depth channel so as to obtain the Depth salient characteristics of the current image; both the RGB salient characteristics and the Depth salient characteristics meet conditional independent distribution and are assumed to obey Gaussian distribution; and performing salient characteristic fusion based on the Bayesian framework to estimate the saliency posterior probability so as to obtain the image saliency area. Therefore, the invention has the beneficial effects that: the high-level saliency characteristics of the RGB image and the Depth image are extracted by using the deep convolutional neural network, the correlation of the saliency characteristics is analyzed, the saliency characteristics are fused under the Bayesian framework and 3D saliency detection is modeled by using the DMNB generation model so as to obtain good accuracy, recall rate and F-measure.

The invention provides an orthotropic dielectric fluid factor and fracture parameter inversion method. The method comprises the following steps: deducing a rigidity matrix of an ortho-symmetric anisotropic fracture medium by using a vertical transverse isotropic medium under a dry rock background according to fracture weakness; with utilization of hypotheses of weak anisotropy and small fracture weakness, deducing a new expression of weak anisotropy approximate rigidity of a saturated fluid rock in an orthogonal medium by combining an anisotropy Gassmann equation; acquiring a PP wave linearized reflection coefficient for fluid and fracture parameter decoupling in the orthogonally symmetrical weak anisotropic medium by combining stiffness disturbance and scattering theories; and with logging information as prior information, realizing elastic impedance pre-stack inversion changing with offset and azimuth angle by using a partial incident angle stacked by azimuth seismic data under Bayesian framework. According to the invention, Reliable results can be provided for fluid identification and fracture characterization; and technical supports are provided for seismic wave propagation research, oil and gas development and seismic disaster prevention.