37 results about "Visual Pattern Recognition" patented technology

First order predicate logics are provided, extended with a bilattice based uncertainty handling formalism, as a means of formally encoding pattern grmmars, to parse a set of image features, and detect the presence of different patterns of interest implemented on a processor. Information from different sources and uncertainties from detections, are integrated within the bilattice framework. Automated logical rule weight learning in the computer vision domain applies a rule weight optimization method which casts the instantiated inference tree as a knowledge-based neural network, to converge upon a set of rule weights that give optimal performance within the bilattice framework. Applications are in (a) detecting the presence of humans under partial occlusions and (b) detecting large complex man made structures in satellite imagery (c) detection of spatio-temporal human and vehicular activities in video and (c) parsing of Graphical User Interfaces.

A media targeting system and method uses visual pattern recognition techniques and wireless device identification in association with consumer transactions as the basis for building a targeting database, which is then later used for automated consumer identification and targeted advertising purposes. The invention solves the problem of generating a large scale, robust media targeting database without relying on active or passive participation by consumers. Also disclosed are other demographics estimation systems and methods which facilitate less expensive media targeting capabilities which can be used in conjunction with the more robust transaction associated method disclosed.

The invention belongs to computer visual pattern recognition and a video picture processing method. The computer visual pattern recognition and the video picture processing method comprise the steps that establishing a three-dimensional space-time sub-frame cube in a video human action reorganization model, establishing a human action characteristic space, conducting the clustering processing, updating labels, extracting the three-dimensional space-time sub-frame cube in the video human action reorganization model and the human action reorganization from monitoring video, extracting human action characteristic, confirming category of human sub-action in each video and classifying and merging on videos with sub-category labels. According to the computer visual pattern recognition and the video picture processing method, the highest identification accuracy is improved by 16.5% compared with the current international Hollywood2 human action database. Thus, the video human action reorganization method has the advantages that human action characteristic with higher distinguishing ability, adaptability, universality and invariance property can be extracted automatically, the overfitting phenomenon and the gradient diffusion problem in the neural network are lowered, and the accuracy of human action reorganization in a complex environment is improved effectively; the computer visual pattern recognition and the video picture processing method can be applied to the on-site video surveillance and video content retrieval widely.

The invention relates to the technical fields of computer vision, mode recognition, biometric identification and the like and particularly relates to a system and method for second-generation ID cardand people verification based on deep learning. The system includes an ID card information reading module, an image acquisition module, a feature value extraction module, a feature value comparison module, and a result display module. The system reads ID card information through the ID card information reading module, collects a scene image through the image acquisition module, extracts feature values of the ID card information and the scene image through the feature value extraction module and compares the similarity of the image feature value, and thus judges the consistency of people and IDcard in a field. A face recognition algorithm based on deep learning is adopted, and the precision and reliability are high.

A method, system and electronic article surveillance tag deactivator detect the presence of an electronic article surveillance tag within a deactivation zone. Video of an item within the deactivation zone is captured. The video is evaluated using a pattern recognition technique to determine the presence of an electronic article surveillance tag within the deactivation zone. The electronic article surveillance tag is deactivated.

The invention belongs to the field of computer vision and pattern recognition, and particularly relates to a cross-domain target re-identification method, system and device based on feature adversarial learning and self-similarity clustering, and aims to solve the problem of poor robustness of an identification result due to the fact that the number of clustering centers is not fixed and the discrimination of feature expression is limited in an existing target re-identification method. The method of the system comprises the following steps: acquiring a to-be-identified image as an input image;for the input image, extracting features of the input image through a pre-trained feature extraction network to serve as first features; and calculating the Euclidean distance between the first feature and the corresponding feature of each image in the image library, sorting the Euclidean distances, and outputting a sorting result. According to the invention, the robustness of cross-domain targetre-identification is improved.

A method and equipment for in-line detecting the flocculated state in water treatment is disclosed. Said method includes such steps as arranging a lamp box and a CCD camera at both sides of a transparent flocculation subsider, taking the optical signals of flocculate, converting them to electric signal, A / D conversion, inputting the digital signals to computer, processing to obtain image information, and analyzing.

The invention belongs to the field of computer vision, pattern recognition and machine learning, particularly relates to a three-dimensional target detection method, system and device based on a self-labeling training sample, and aims to solve the problems that real labeled data is difficult to obtain and high in cost, and a model trained by virtual data cannot adapt to a real scene. The method comprises the steps of: performing three-dimensional target detection of an input image sequence through a trained model, wherein the model training method comprises the steps that a high-quality modelis embedded into a CARLA simulator; enhancing a point cloud data sample generated by the CARLA simulator through a sampling algorithm based on laser radar guidance; and on the basis of a three-dimensional target detector VoxelNet, performing domain offset alignment by introducing domain self-adaptive modules of a voxel level and an anchor point level, and adding consistency constraints to build adomain self-adaptive three-dimensional target detector DA-VoxelNet. According to the invention, the three-dimensional target detection model trained by the virtual data can adapt to a real scene, thedetection effect is good, and the precision is high.

A method of mid-level pattern recognition provides for a pose invariant Hough Transform by parametrizing pairs of points in a pattern with respect to at least two reference points, thereby providing a parameter table that is scale- or rotation-invariant. A corresponding inverse transform may be applied to test hypothesized matches in an image and a distance transform utilized to quantify the level of match.

The invention relates to a method and system for regularly detecting an X-ray machine image resolution testing body, and belongs to the technical field of image resolution. The method includes the following steps of S1, image testing, S2, zone locating, S3, TEST1 detecting algorithm testing, S4, TEST2 detecting algorithm testing, S5, TEST3 detecting algorithm testing, and S6, TEST4 detecting algorithm testing. The method and system for regularly detecting the X-ray machine image resolution testing body intends to integrate leading edge theories of compute vision and mode recognition and develops an automatic regular checking technology of an X-ray security checking machine based on intelligent image analyzing, and through automatic analyzing of imaging of a standard testing box, whether the performances of the line resolving power, penetrating resolving power, space resolving power, penetrating power and the like of the X-ray security checking machine are up to standard or not are accurately and objectively diagnosed in real time to give expert-level opinions on whether security devices can be normally used or not.

The invention provides a pulmonary nodule false positive screening method based on multi-task learning. The method comprises the following steps: 1, acquiring an initial CT image of a lung to be detected through a CT detector, and preprocessing the three-dimensional initial CT image to obtain a standard CT image; 2, slicing the standard CT image, and inputting a CT slice with a fixed size into a candidate nodule detection model to obtain a candidate nodule; 3, marking positive nodules and false positive nodules of the candidate nodules obtained in the step 2, and inputting the marked nodules into a multi-task model to obtain benign and malignant classification of the candidate nodules and a reconstructed input image; the method is characterized in that a doctor is helped to complete auxiliary diagnosis work by utilizing computer-aided detection and utilizing cross technologies such as digital images, computer vision and pattern recognition. By means of the computer-aided diagnosis technology, the suspected nodule area can be detected from the CT image quickly and accurately, the workload of the doctor can be reduced, and meanwhile, the film reading accuracy and efficiency can be improved.

The invention discloses a machine vision based forward vehicle collision warning system and a method for realizing front vehicle collision warning by the system. An advanced machine vision pattern recognition algorithm is adopted, a high-speed digital signal processor is combined to analyze traffic road conditions acquired by a forward vision camera mounted on a vehicle in real time, a warning in the vision, audition or touch aspect is given out to a driver on the condition that the vehicle is predicted to be about to collide with a forward vehicle so as to remind the driver of safety driving, and accordingly rear-end collision is reduced effectively.

The invention provides a target tracking method based on multilayer time sequence filtering, and relates to the technical field of computer vision mode recognition. The method comprises the followingsteps: step 1, selecting and determining a target object to be tracked from an initial image, step 2, a multi-layer time sequence filtering network comprising a basic time sequence network 1 and a basic time sequence network 2 which have the same network structure and are parallel to each other; step 3, carrying out multi-layer time sequence filtering network training, adopting an Adam optimization method for training, wherein the multi-layer time sequence filtering network has a target positioning capability; step 4, extracting a video image as an input image to be tracked; extracting frame images one by one as input images according to a time sequence; step 5, performing preliminary estimation on the target position through the basic time sequence network 1; and taking the initial imagein the step 1 as Ft, and taking the currently input frame image as Ft+1; step 6, accurately positioning the target through the basic time sequence network 2.

The invention discloses a GMPHD video multi-target tracking method fusing correlation filtering, and belongs to the technical field of computer vision, mode recognition and information processing. According to the method, a target is tracked by adopting a related filtering thought, and intersection ratio judgment of image information is added to carry out target template updating-free and parameter processing on the shielded target, so that pollution of a target template is reduced, and mistaken following frames are reduced. According to the method, a shielded target is put into Gaussian mixture probability hypothesis density filtering to perform position prediction updating operation, and if the target reappears in the later period, the target label is associated again, so that fragmentedtracks are reduced, and the defect of missing detection of a detector is overcome. Finally, a result on an MOT17 data set proves that compared with the best tracking algorithm GMPHDOGM17 related to GM-PHD at present, the multi-target tracking accuracy MOTA index is improved to 50.3 from the original 49.9.

The invention provides a traction substation outdoor insulator abnormity detection method. The invention relates to the technical field of computer vision, pattern recognition and intelligent systems.The method comprises the steps of: respectively constructing data sets of an insulator positioning network and an insulator image generation network; constructing an insulator positioning network, and enabling the network to obtain the capability of positioning the insulator in the image through training; constructing an insulator image generation network, and obtaining the insulator image reconstruction capability through training; inputting the traction substation image into a network model; positioning the insulator through an insulator positioning network, and extracting an insulator image; and carrying out anomaly detection on the insulator, and giving an anomaly score to each picture by the insulator image generation network; setting an abnormality judgment threshold value, if the abnormality score exceeds the set threshold value, judging that the sample is an abnormal sample, and if the abnormality score is lower than the threshold value, judging that the sample is a normal sample; and finally, performing feature extraction on the judged abnormal image and the generated image thereof, and comparing the difference to locate an abnormal area.

The invention belongs to the field of computer vision and pattern recognition, particularly relates to a pedestrian search method, system and device based on a bidirectional interaction network, and aims to solve the problem of low pedestrian search accuracy caused by the influence of scene information outside a bounding box on pedestrian features. The method of the system comprises the followingsteps: acquiring a tth frame image in an input video as an input image; normalizing the input image, and obtaining a search result of a pedestrian target to be searched through a pre-trained pedestrian search model; constructing the pedestrian search model based on a residual network, and adding a new classification layer in a classification regression layer of the residual network and is used forobtaining the identity classification probability of a pedestrian target. The accuracy of pedestrian search is improved.

A computer system, non-transitory computer storage medium, and a computer-implemented method of assessing technical risk using visual pattern recognition in an Information Technology (IT) Service Management System. A data visualization engine and a time series generation engine receive the operational data, respectively. A first representation of the data is generated by the data visualization engine, and a second representation of the data is generated by the time series generation engine. Anomaly patterns are identified by a pattern recognition engine configured to perform feature extraction and data transformation. An ensembler is configured to accept the outputs from two AI anomaly engines and make a final decision of whether anomaly patterns are captured. Risk scores based on the identified anomaly patterns are output by a pattern recognition engine to an automated management system. The anomalies includes information regarding vulnerabilities of devices or components of the IT Service Management System.

The invention belongs to the technical field of digital image processing, and relates to relevant theoretical knowledge such as computer vision, pattern recognition, machine learning and data mining. The present invention first uses the optical flow histogram to extract the optical flow features of two adjacent frames of pictures, and extracts the gradient histogram information of a single frame image to obtain the motion feature information in the monitoring scene, adopts the feature information of low-dimensional space and follows the [action 1|Action 2|Action 3|...]; then use K-means to cluster, and after obtaining the cluster center, use the cluster center as an over-complete dictionary to solve the sparseness of the test sample under the over-complete dictionary rank expression to obtain the expression matrix; finally, according to the maximum value in the expression matrix, the behavior category to which the test sample belongs is calculated. The invention adopts low-rank and sparse human body action recognition and cross-validation method, the recognition rate is 92.3-98.79%, and the misrecognition rate is 1.21-7.6%. The invention has the characteristics of low rank, and the recognition rate reaches 92.3-98.79%, and the false recognition rate is 1.21-7.6%.

The invention belongs to the fields of computer vision, pattern recognition and intelligent transportation, and specifically relates to a distributed license plate recognition method, system and device based on multi-attribute fusion, aiming to further improve the accuracy, effectiveness and credibility of license plate recognition results in natural scenes Spend. The system method includes obtaining the license plate image to be recognized; extracting the features of the license plate image as initial features; performing deep encoding on the initial features to obtain the type features related to the type of the license plate image, and obtaining the type prediction result through the image type classifier; The initial features are deeply encoded to obtain the color features related to the license plate color, and the color prediction result is obtained through the image color classifier; the initial features, type features, and color features are fused to perform deep encoding to obtain the text features related to the license plate image. And get the license plate number recognition result through the pre-built character sequence generator. The invention improves the recognition accuracy, effectiveness and reliability.