140 results about "Fuzzy graph" patented technology

Described is a behavior recognition system for detecting the behavior of objects in a scene. The system comprises a semantic object stream module for receiving a video stream having at least two frames and detecting objects in the video stream. Also included is a group organization module for utilizing the detected objects from the video stream to detect a behavior of the detected objects. The group organization module further comprises an object group stream module for spatially organizing the detected objects to have relative spatial relationships. The group organization module also comprises a group action stream module for modeling a temporal structure of the detected objects. The temporal structure is an action of the detected objects between the two frames, whereby through detecting, organizing and modeling actions of objects, a user can detect the behavior of the objects.

The invention discloses an edge information-based multi-scale blurred image blind restoration method, which comprises the following steps of: circularly and gradually restoring an image from a small scale layer to a large scale layer, setting self-adaptive parameters at different scales, and processing each scale layer, namely bilaterally filtering the restored image to obtain an image of which the noise and ripple are removed; performing shock wave filtering to obtain an image with high-strength contrast edges; solving the edges, and combining a fuzzy core initial value and a fuzzy graph to obtain an accurate fuzzy core; restoring a fuzzy image at the current scale to obtain a clear restored image by using the solved fuzzy core; sampling and amplifying in the current scale layer to obtain the restored image and a fuzzy core initial value of an adjacent large scale layer, and performing cycle operation on the adjacent large scale layer. The edge information-based multi-scale blurred image blind restoration method can effectively converge various images in different fuzzy degrees, and compared with a general blind restoration method which directly solves the energy minimization, the blurred image blind restoration method has the advantages of low computational complexity and high noise suppression capacity.

The invention relates to an automatic lip gloss image enhancement method based on the color space. The method comprises the following steps that 1, face identification and five sense organ locating are carried out on an image, and a lip contour area is determined; 2, fuzzy processing is carried out on the lip contour area, and a lip contour fuzzy graph is generated; 3, according to a probability graph of the color space, the probability that each pixel in the lip contour area is the lip is calculated, a lip probability graph is generated and is combined with the lip contour fuzzy graph generated in the step 2, and then a final probability graph is obtained by calculation; 4, according to the final probability graph and the lip gloss color selected by a filter, each pixel in the lip contour area is automatically coated with lip gloss, and finally a result graph after being automatically coated with the lip gloss is obtained. Compared with a lip gloss processing method in the prior art, the method has the advantages of being easier to implement, higher in speed, better in identification precision and more applicable to mobile intelligent equipment.

The invention relates to a sparse representation-based blind restoration method of a broad image, aiming at solving the technical problem that an image restored by the existing broad image blind restoration method is bad in effect. The method has the technical scheme that the broad image is sparsely represented by a fuzzy dictionary and a brilliant image is rebuilt by a brilliant dictionary to restore the image due to the characteristic that the sparse coefficient of the broad image under the fuzzy redundancy dictionary is coincident with that of the brilliant image under the brilliant redundancy dictionary. The illcondition during deconvolution can be avoided, the ringing effect at a high edge during image restoration can be reduced, and the image which is more brilliant can be obtained.

The invention discloses a defocus blurred image sharpness processing method, which is implemented according to the following steps: firstly, the mean value and the variance of the edge width is calculated according to gradient information of an image; an initial parameter of a fuzzy model is obtained according to the statistical data; secondly, the blurred image is divided into concentric circles around the center of the image and is broken down into sub-images, wherein, the number of the concentric circles is k and the number of the sub-images is k+1; the corresponding blurred initial semidiameters are distributed to the sub-images; the optimal blurred semidiameter is found through an iterative mode, and the sharpness processing of all sub-images is performed by adopting a frequency domain inverse-filtering manner; finally, all the sub-images are added to synthesize the whole sharpness image. The defocus blurred image sharpness processing method overcomes the limitations in the existing circular disc function modeling recovery method based on the fixed semidiameter, and serves the purpose for restoring the sharp image.

The invention belongs to the field of image restoration, and particularly relates to a fuzzy kernel refining-based blind simple image motion blurring removal method. The method mainly comprises threesteps of: 1, importing an effective strong edge to carry out multiscale fuzzy kernel estimation, and outputting a fuzzy kernel estimation value k and a clear image estimation value I' of each scale according to an input fuzzy image B; 2, forming fuzzy kernel post-processing by hard threshold value processing, connectivity verification and morphological closed operation, and carrying out fuzzy kernel post-processing on the fuzzy kernel estimation k of the highest scale; and 3, carrying out Lapras non-blind convolution removal, and outputting a final fuzzy kernel kR and a final clear image estimation value If. Aiming at the defect that fuzzy kernel estimation is incorrect and is not sparse and continuous enough, the method imports the effective strong edge and the fuzzy kernel post-processing, so as to carry out effective estimation on fuzzy kernels with various forms and various scales and then obtain de-blurring results which are remarkable in effect and extremely close to real clear images.

Described is a behavior recognition system for detecting the behavior of objects in a scene. The system comprises a semantic object stream module for receiving a video stream having at least two frames and detecting objects in the video stream. Also included is a group organization module for utilizing the detected objects from the video stream to detect a behavior of the detected objects. The group organization module further comprises an object group stream module for spatially organizing the detected objects to have relative spatial relationships. The group organization module also comprises a group action stream module for modeling a temporal structure of the detected objects. The temporal structure is an action of the detected objects between the two frames, whereby through detecting, organizing and modeling actions of objects, a user can detect the behavior of the objects.

The present invention relates to a blurred image recognition method and apparatus based on an SIFT algorithm. The method comprises: respectively carrying out Gaussian smoothing, gray level calculation and motion blurring on clear images to generate a fuzzy space; carrying out feature extraction and description on all the images in the fuzzy space based on the SIFT; extracting feature points of a blurred image to be identified by utilizing the SIFT algorithm; respectively matching the feature points of the image to be identified with feature points of each image in the generated fuzzy space; and carrying out estimation on all the matching based on a preset estimating standard, screening an image with the optimal matching effect and using the image with the optimal matching effect as the final recognition result.

The embodiment of the invention provides a driver attribute identification method and a related product, and the method comprises the steps: obtaining an input image which comprises a driver area image; extracting a driver area image in the input image; determining a fuzzy region and a clear region in the driver region image; performing defuzzification processing on the fuzzy area based on a preset maximum posteriori framework model to obtain the processed driver area image; and performing attribute identification on the processed driver region image by adopting a preset deep neural network model to obtain a target attribute identifier corresponding to the driver. According to the method and the device, the blurred image can be clearly processed, so that the identification precision of thedriver attribute is improved.

The invention provides a feature reconstruction layer training method, an image feature reconstruction method and related devices, and relates to the field of pedestrian re-identification. The training method is applied to the electronic device and comprises the steps of obtaining a first feature vector according to a training image set; wherein the training image set comprises at least one training image; obtaining a second feature vector according to the fuzzy training set; wherein the fuzzy training set comprises at least one first feature map, and the first feature map is obtained by performing image feature suppression on the training image; obtaining reconstruction layer updating information according to the first feature vector and the second feature vector; and obtaining a featurereconstruction layer matched with the model convergence condition according to the reconstruction layer updating information. By use of the feature reconstruction layer obtained by the invention, no matter whether the image to be retrieved is a clear image or a blurred image, a better image feature can be obtained so as to reduce interference and influence of image blurring and the like on neuralnetwork features, and then pedestrian retrieval accuracy is improved.

The invention relates to a restoration method for a blurred image caused by camera shaking. In the restoration process of the blurred image caused by camera shaking, if a fuzzy model is designed inappropriately, a restoration result can be bad, computation efficiency is low, and internal storage footprints are increased suddenly, so that it is a problem demanding prompt solution in the field to develop the better restoration method for the blurred image caused by camera shaking. A generalized additive convolution model is designed, and the blurred image caused by camera shaking is restored based on the model. The restoration method includes the steps that motion paths of camera shaking are estimated; all slice-shaped paths and fiber-shaped paths are calculated, and proportions for which the slice-shaped paths and the fiber-shaped paths account for are set through a greedy algorithm; non-blind restoration is performed through an APG algorithm based on mixed GACM. The restoration method is good in restoration visual effect, small in internal storage footprint and suitable for restoring various blurred images caused by camera shaking and takes efficiency and speediness into account.

The invention discloses an image deblurring method and system based on deep neural network parameter estimation, and the method comprises the steps: obtaining a training set and a test set, and carrying out the preprocessing; setting network parameters; performing gaussian blur removal on the image based on deep neural network parameter estimation, wherein the deep neural network comprises two sub-modules, namely a Gaussian standard deviation level parameter estimation sub-module and a non-blind deblurring sub-module, the Gaussian standard deviation horizontal parameter estimation sub-module is of an hourglass type network structure, a skip connection mechanism is used between a decoding block and a coding block in a symmetric layer; carrying out PCA principal component analysis on a Gaussian blurred kernel, then carrying out dimension stretching to obtain a vector graph, and taking the vector graph and a blurred image as input of the non-blind deblurring module; enabling the non-blinddeblurring sub-module to execute a non-linear mapping process by applying cascaded convolution layers; and training a neural network, and testing the neural network to obtain a deblurring result. Themethod is applied to image deblurring, and a good effect can be achieved.

The invention provides a deep learning-based high-precision image fuzzy detection method, which comprises the following steps: (1) building a deep convolution neural network model CNN and carrying out initialization and inputting a detected image into the deep convolution neural network model CNN; (2) selecting different s scales for a to-be-detected image by the deep convolution neural network model to obtain image blocks of different scales; (3) carrying out feature extraction on the image blocks in the step (2) by the deep convolution neural network model to obtain single-scale fuzzy graphs according to six convolution layers; and (4) carrying out fusion processing on the different single-scale fuzzy graphs by the deep convolution neural network model for multiple times to obtain the fuzzy graphs. According to the method, a deep convolution neural network is applied to the problem image fuzzy detection, so that a fuzzy area in the image is accurately detected as the target.

The invention belongs to the technical field of digital image intelligent processing, and particularly relates to an image defocusing blurring method based on deep learning. The method comprises the following steps: constructing a defocusing blurred data set by shooting or adding random blurring and the like, so that each group of data comprises a clear image as an original image and a plurality of blurred images as blurred images corresponding to the clear image; training a defocusing fuzzy deep neural network; recovering a blurred object out of a focal plane from the image through a deep neural network by using a non-alignment loss function; a non-pixel-level aligned deblurred data set is shot in a real scene, and a deep neural network is trained through a non-alignment loss function. Experimental results show that the out-of-focus blurred image shot in a real scene can be effectively recovered, and the proposed data set can effectively train the out-of-focus blurred network througha non-alignment loss function. The method can be used for camera zooming, robot vision systems and the like.

The invention discloses a blind image deblurring method based on depth prior. A deep convolutional neural network DIP-Net is used for implicitly modeling an image smoothness prior constraint to generate a clear image; estimating a fuzzy kernel by solving an accurate solution about a fuzzy kernel optimization problem; and alternately iteratively updating the blurring kernel and the clear image, calculating a loss function by using the restored clear image and the blurring kernel, and updating network parameters. Carrying out joint modeling on the blurred image and the blurred kernel, and simultaneously estimating a clear image and the blurred kernel by adopting a mode of alternately iterating a network model and a mathematical model; blind deblurring of end-to-end self-supervised learning is achieved only using blurred images without any additional implicit or explicit image priors. The regularization method is realized in combination with a deep network structure, and a blurred image and a blurred kernel truth value training network do not need to be used; compared with a traditional model method, the method does not need to employ an image pyramid mode to estimate the blurring kernel from coarse to fine, and effectively inhibits the noise in the restored image.

The invention discloses an optical aberration blur removing method based on deep learning. The method comprises the following steps: 1) obtaining a point spread function of the optical system with aberration; 2.1) selecting a high-resolution image, and performing energy domain transformation to obtain an energy domain image; 2.2) performing block convolution on the energy domain image by using thecalculated correction point diffusion matrix to obtain an energy domain simulation fuzzy graph; 2.3) performing numerical domain transformation on the energy domain simulation blurred image to obtainan aberration blurred image, and forming an aberration blurred data set; 3) based on the aberration fuzzy data set, training an aberration correction neural network; and 4) correcting an image shot by the aberration optical system developed and produced by using the optical parameters through the aberration correction neural network obtained by training in the step 3) to obtain a corrected image.When the method is used, optical parameters of a camera (camera head) are operated by adopting the method disclosed by the invention, and image blurring caused by aberration of an optical system canbe well eliminated.

The invention provides an intelligent method and system integrating deep learning and logical judgment. The method comprises the following steps: acquiring an answer sheet image; performing image segmentation on the answer sheet image by using a pre-trained image segmentation model so as to identify each option area and parameter information thereof on the answer sheet image, the parameter information including coordinate information and size information; according to the coordinate information of the option area, judging whether a missing option area exists or not; if yes, supplementing the missing option area according to the coordinate information of the adjacent option area of the missing option area and the size information of the option area; and outputting each option area of the identified answer sheet image. By fusing the deep learning neural network model and the logic judgment, the method successfully solves the problems that the fuzzy image feature points are few, the image noise is large and the image noise is difficult to process, improves the identification precision of the option region, and can well solve the problem of missed identification.

The invention brings forward a detection identification method of a standard constant temperature bath glass thermometer. The method comprises the following steps: step 1, obtaining a fuzzy image of a character to be identified in the glass thermometer; step 2, performing contour fuzzy feature extraction on the character to be identified; and step 3, performing fuzzy closeness identification processing on the fuzzy image of the character to be identified so as to identity the character. According to the technical scheme provided by the invention, incomplete characters on the glass thermometer are identified by use of a fuzzy closeness algorithm which is brought forward lately, the identification correct rate is about 96%, and the thermometer reading accuracy can reach 95%.

The invention discloses a missile-borne image deblurring method based on a generative adversarial network, and belongs to the technical field of missile-borne computer vision. According to the method, a fuzzy missile-borne image is deblurred by using a generative adversarial network, a deep convolutional generative adversarial network model comprising a generator and a discriminator is designed, the generator adopts a coding-decoding structure, a joint loss function is constructed, and continuous training is performed to generate a restored image of the missile-borne blurred image. A distinct image and an image forged by a generator are distinguished through a discriminator, the generator approaches the distinct image confusion discriminator, a network model reaches an expected index through adversarial training of two networks, and the network model after adversarial training is transplanted to a missile-borne computer for deblurring a missile-borne image and improving guidance precision. Besides, different fuzzy sources are simulated by establishing a vivid semi-physical simulation system for synthesizing motion blur, the problem that actual acquisition of missile-borne image data is difficult is solved, the training efficiency is improved, and the test cost is saved.

The invention relates to a blurred image restoration method based on an unsupervised generative adversarial network, and the method comprises the steps: employing two generators GA and GB to carry out the modeling of a restoration problem of a blurred image and a blurring degradation process of a clear image under the framework of a dual generative adversarial network; introducing a discriminator to cooperate with a corresponding generator to carry out adversarial training, wherein a discriminator DA is used to determine whether an image generated by the GA is clear, and a discriminator DB is used to determine whether an image generated by the GB is fuzzy; meanwhile, introducing L2 pixel reconstruction loss and perception loss to carry out modeling on a target loss function; finally, training the network by using a non-paired fuzzy-clear image data set, updating network parameters by using an Adam optimizer, and finally obtaining a local optimal solution of the model. Compared with a traditional blurred image restoration method, the method has the advantages that the workload of data set construction is greatly reduced, and the method has wide practical value and application prospect.

The invention relates to a defocusing fuzzy kernel estimation method based on binocular stereoscopic vision. The method comprises the following steps: calculating an initial fuzzy kernel; acquiring anequivalent blurred image; carrying out three-dimensional matching calculation: synthesizing the gray information and the fuzzy kernel information as data items of an energy function of a global matching method, and performing three-dimensional matching on the same fuzzy image by optimizing the energy function by using the global matching method to obtain a dense disparity map; disparity post-processing: performing weighted mean filtering processing on the dense disparity map; and performing final fuzzy kernel calculation: taking the disparity map subjected to disparity post-processing as an initial disparity map, and performing fuzzy kernel calculation again to obtain a final fuzzy kernel. According to the method, the relation between parallax and the fuzzy kernel in binocular stereoscopic vision is fully considered, the accuracy of fuzzy kernel calculation is improved through the binocular image, and the method is suitable for defocusing fuzzy kernel estimation based on binocular stereoscopic vision.

The invention discloses the construction of an interval hesitant fuzzy graph decision-making method considering relevance and a priority relationship, wherein model establishment and model calculationresearch of the interval hesitant fuzzy graph multi-attribute decision-making method considering relevance and priority relation are included, the interval hesitant fuzzy graph of relevance between attributes is described by inputting an interval hesitant fuzzy decision-making matrix and a linear priority relation between the attributes, and an optimal alternative scheme is outputted. The specific calculation comprises the steps of calculating an interval hesitant fuzzy information energy coefficient psi ij between the attributes ai and aj with relevance of an alternative decision scheme; calculating an attribute weight omega j by using the linear priority relationship among the attributes; and calculating the overall attribute value of the alternative decision-making scheme pk (k = 1, 2,..., m), and calculating the score value of the overall attribute value corresponding to the alternative decision-making scheme pk to determine the optimal alternative decision-making scheme. According to the present invention, by utilizing the characteristics and the advantages of the interval hesitant fuzzy graph, an effective solution is provided for a complex uncertain multi-attribute decisionproblem with relevance and the priority relationship.

The embodiment of the invention provides an image deblurring method and device and electronic equipment. The method comprises the steps that a target blurred image to be deblurred and depth information corresponding to the target blurred image are acquired; determining a target depth value of the target blurred image based on the depth information corresponding to the target blurred image, the target depth value being used for representing an average depth of each pixel point contained in the target blurred image; determining a target blurred kernel parameter corresponding to the target depthvalue based on a pre-constructed corresponding relationship between the depth value of the blurred image and the blurred kernel parameter; generating a target fuzzy kernel with a target fuzzy kernel parameter; and performing a deconvolution operation on the target blurred image by using the target blurred kernel to obtain a deblurred image corresponding to the target blurred image. Through the technical scheme provided by the embodiment of the invention, the efficiency of determining the blurred kernel of the blurred image can be improved, and the deblurred image corresponding to the blurred image can be quickly obtained.

The invention discloses a motion blur restoration method based on a contour enhancement strategy, and relates to the technical field of image processing, and the method comprises the following steps:(1) encoding an original blur image, removing noise information through fine adjustment of a restoration network, and finally restoring a clear contour image through a decoder; respectively carrying out contour extraction on the image by using a Sober operator and a Canny operator, forming a sequence with the contour recovered by the decoder, carrying out further restoration extraction on the image contour by using LSTM, and finally generating a sharp edge; (2) respectively sampling and encoding the original fuzzy graph and the sharp edge, pairing the graph codes and the sharp edge codes withthe same size one by one, and outputting graph code and sharp edge code pairs; (3) generating a potential clear graph by using a multi-scale repair framework; according to the invention, the sharp edge in the image with serious motion blur can be extracted, so that the generated sharp edge effectively assists a multi-scale framework in removing the motion blur, and the efficiency of removing the motion blur is effectively improved.

The invention relates to an image processing technology, and discloses a method for identifying a blurred image, which comprises the following steps of: generating disturbance data for a training image; synthesizing the disturbance data and the training image to obtain a training blurred image; utilizing the training blurred image to train an image blurring discrimination model to obtain a first image blurring discrimination model and a training prediction result; calculating an error value between the training prediction result and a preset fuzzy label; collecting the training blurred imagesof which the error values are greater than an error threshold value into an error sample training set; training the first image fuzzy discrimination model by using the error sample training set to obtain a second image fuzzy discrimination model; and inputting the to-be-judged image into the second image fuzzy judgment model to obtain a judgment result. The invention provides a method and device for identifying the blurred image, equipment and computer readable storage medium In addition, the invention also relates to blockchain technology, and the to-be-judged image can be stored in a blockchain node. According to the invention, the efficiency and accuracy of identifying the blurred image can be improved.

The invention discloses a robot camera calibration method based on edge scale adaptive defocusing fuzzy estimation, which comprises the following steps: firstly, acquiring a checkerboard picture by using a camera to be calibrated, and carrying out Canny edge detection and corner detection on the acquired checkerboard picture; secondly, constructing an edge graph with a consistent scale according to a Canny edge detection value; then, setting a local scale value of edge detection while setting blurring values of two-time re-blurring images; carrying out gaussian blur on the checkerboard image again, and solving the gradient ratio of the two re-blurred images; then, calculating the defocusing fuzzy quantity of each corner point of the checkerboard original picture; drawing a circle by taking the detected angular point as a circle center and the defocusing fuzzy quantity as a radius, and setting a weight of a camera calibration energy equation; and finally, according to the optimized camera calibration energy equation, performing iteration on the energy equation in the obtained circle range until convergence, and outputting an optimal camera calibration parameter, so that the camera calibration precision can be greatly improved.

The invention discloses a method for measuring vehicle speed based on a fuzzy image. The method for measuring the vehicle speed based on the fuzzy image comprises the following steps that step 1, a single vehicle scene image is shot in real time; step 2, a calibration function between road surface distance and pixel distance is calculated according to the vehicle scene image; step 3, a vehicle image is extracted from the current vehicle scene image; step 4, fuzziness calculating is carried out on the vehicle image, and a fuzzy pixel value is output; and step 5, the vehicle speed is calculatedaccording to the calibration function, the fuzzy pixel value and shooting exposure time. According to the method for measuring the vehicle speed based on the fuzzy image, the single image is used forcalculating the vehicle speed in the image by a deep learning model evaluating the fuzziness, and the installation and measurement process of vehicle speed measuring equipment can be effectively simplified.

The invention discloses a calculation method for automatically zooming content sensitive image with prominent depth-of-field characteristics, includes A1 selecting a plurality of patches arbitrarily from 1000 high-definition images, A2 obtaining a fuzzy dictionary Dblur by training ((using Gaussian fuzzy of sigma=2), A3 each patch in the target image G being then decomposed into a weighted superposition sum of several basic atoms, A4 by calculating the number of basic atoms decomposed by the target patch, clear images having a lot of detail, more atoms in Dblur being needed to approximate, theblurred image being decomposed into fewer atomic superpositions, A5 describing the number of atoms fblur in the image decomposition as the fuzzy depth of field and proposing a depth of field estimation algorithm, an A0 depth of field estimation algorithm being obtained; A6 reducing the distortion of non-salient region by enhancing the protection formula of background in x-direction and y-direction for depth of field F, and A 7 keeping salient region and reducing the change of structure information of non-salient region, which can alleviate the distortion of non-salient region.