278results about How to "Solving recognition problems" patented technology

The invention discloses a pedestrian re-identification method based on multi-attribute and multi-strategy fusion learning. The method of the invention includes the steps of in an offline training phase, firstly selecting pedestrian attributes which are easy to be judged and have a sufficient distinguishing degree, training a pedestrian attribute identifier on an attribute data set, then labeling attribute tags for a pedestrian re-identification data set by using the attribute identifier, and next, by combining the attributes and pedestrian identity tags, training a pedestrian re-identification model by using a strategy fused with pedestrian classification and novel constraint comparison verification; and in an online query phase, extracting features of a query image and images in a database by using the pedestrian re-identification model, and calculating the Euclidean distance between the feature of the query image and the feature of each image in the database to obtain the image with the shortest distance, which is considered as the result of pedestrian re-identification. In terms of performance, the features in the invention are distinguishable and high accuracy is obtained; and in terms of efficiency, the method of the invention can quickly search for the pedestrian indicated by the query image from the pedestrian image database.

An accurate high-speed moving vehicle mark identification method based on video belongs to a calculation processing method adopting image and mode identification technology to realize vehicle detection and vehicle mark identification. The present invention mainly comprises five function modules, namely a video camera control module, a motion detection module, a vehicle snapshot module, a vehicle positioning module and a vehicle identification module. The present invention acquires a real-time video image through a camera system, judges whether a vehicle passes by processing the video image, judges the motion direction of the vehicle according to motion characteristics, segments a picture of the moving vehicle from an image sequence, positions a vehicle mark through vehicle texture features, and performs classification identification to a vehicle head mark or a vehicle back mark by utilizing vehicle mark features before and after. The present invention can solve the vehicle model and vehicle mark identification of high-speed running vehicles, especially the identification of vehicle back marks.

The invention relates to smart agreement analytical methods and devices used for intruding detection defense (IDS / IPS) and audit products. The purpose of the invention is to provide an agreement not to rely solely on the static ports and matching agreement characteristics of intelligent field protocol analysis technology and analytical format of the agreement is automatically adjusted in different versions of the software and gives accurate results, which enhanced the accuracy of the analysis of the agreement. The invention consists of three major steps: the establishment of agreements features model; agreement recognition; intelligent analysis of that agreement. This invention solved the traditional IDS / IPS products for the non-standard ports or did not have static characteristics of field data packet network protocol identification of problems but for some applications or different versions of the agreement, such as the reasons for the analytical results can provide automated error rectification work.

The invention relates to a distributed vision computing method based on an open type Web Service framework. The method has the technical scheme of: collecting the video image of an identified target by adopting an embedded video collecting front end, and sending the video image back to a data center through network multicast; a computing server is controlled by a control server in the data centerto accomplish the computing task, and carrying out the uncompressing, target detecting, image preprocessing, feature extraction, identification, classification and result drawing of video data; and integrating the computing results. The method solves the defect of difficult compatibility with other systems. Under the uniform scheduling of the control server, the method avoids a great deal of unnecessary matching calculation, improves the performance of system operation, achieves the compatibility with most of databases and simply updated external systems, and realizes the identification and calculation of large-scale databases by adopting a packet mode.

The invention discloses a recognition and counting method for cells, comprising nine steps: preprocessing an image in a micron-order microscopic acquisition environment; extracting cell holes from the preprocessed image; performing closed hole filling of cells by using the knowledge of a connected domain; extracting a contour point sequence of cells from the image filled in step 3; filling non-closed holes of cells by adopting a non-closed hole filling method based on circularity determination; performing chamfer distance transformation on the filled image; performing extreme value uniqueness marking on cell hole positions; segmenting the image after extreme value uniqueness by using a marked watershed method; and quantifying and marking the segmented result. The method has the advantages that the influence of image noise can be greatly reduced, the phenomena of over-segmentation and discontinuous segment lines are eliminated, the segmentation effect is improved, and the cell recognition rate is improved.

The invention discloses an entity relationship joint extraction method based on multi-label labeling and a composite attention mechanism, which comprises the following steps: collecting corpus data for research, then removing sentences of which the relationship labels are 'None', and performing multi-label labeling on the rest sentences to form a training set; inputting the sentences subjected tomulti-label labeling into a joint extraction model, identifying entities contained in the sentences and relationships among the entities through the joint extraction model, and constructing a triple;and correcting the extracted triples by utilizing a relationship alignment model so as to adapt to multi-label labeling of (head entity E1 and tail entity E2) entity pairs. The method has the advantages that the accuracy of triple extraction can be effectively improved, and the method is an effective tool for information extraction of unstructured data.

The present invention belongs to the image character recognition field and relates to a recurrent neural network-based complex image character sequence recognition system. The system includes an image character input module, a slide sampling module, a CNN and an RNN, wherein the image character input module is a scanner, a digital camera or an image character storage module. The slide sampling module in the system performs sliding sampling on an image character sequence to be recognized and inputs sampled sub pictures into the CNN; the CNN extracts features and outputs the features to the RNN; and the RNN recognizes the front part of a Chinese character, the back part of a Chinese character, numbers, letters or punctuations according to the input signals of the CNN, the output data of the CNN at the last time point, and vector data converted from the recognition result of the CNN at the last time point. With the system of the invention adopted, problems in the segmentation of a complex image character sequence can be solved, a language model is not required to be constructed additionally, and the recognition efficiency and accuracy of the complex image character sequence can be significantly improved.

The invention relates to an unknown network protocol identification method. The method includes the steps: firstly, taking network data packets as input, and representing each network data packet as a characteristic vector which can be used for classification; secondly, taking the obtained characteristic vectors as input to form a characteristic vector data set, and using a support vector machine oriented active learning method for learning the characteristic vector data set to obtain a classifier aiming at a to-be-tested network protocol; and thirdly, using the obtained classifier to discriminate protocol attributes of to-be-identified network data packets. The invention further provides an unknown network protocol identification system which comprises a data packet modeling module, a classifier construction module and an identification module. The active learning method is adopted, few tagged samples can be used for achieving excellent learning efficiency, and accordingly, the number of the tagged samples in the learning process is reduced effectively, and analyzed network protocols can be identified accurately from miscellaneous network flows.

A layered coordinate network virus and vicious code recognizing method. The characteristics are: use strong self-protection mechanism of biological immunity for reference, correspond the network virus and vicious code recognizing method to the multi-level protective mechanism of biological immunity system, judge the dangerous degree of the stand-by detected script through statistically analyzing the word frequency of the key words, based on the point view of 'self-collection' of register form operation analyze and judge the exceptional behavior of the register form written in the form route, and recognize non-self the executing sequence of the programming interface of the applied program, at last send all the exceptional behavior information to the network control station through the network. It well solves the problem of the identification of the unknown network virus and vicious code, and has good capacity of identification, realizes the monitoring and management of the network virus and vicious code of single system and the whole sub-network.

The present invention provides a voice recognition method and device. The voice recognition method comprises the following steps: a voice command is received by a mobile terminal, and the intensity of a first network signal of the mobile terminal is obtained; when the intensity of the first network signal is not less than a preset threshold, the voice command is sent by the mobile terminal to a cloud server, and the intensity of a second network signal of the mobile terminal is obtained; when the intensity of the second network signal is less than the preset threshold, the voice command is compared with the voice command prestored in the mobile terminal through the mobile terminal to determine whether the voice command is matched with the voice command prestored in the mobile terminal or not; and if the voice command is matched with the voice command prestored in the mobile terminal, a first voice recognition result may be outputted by the mobile terminal according to the prestored voice command. According to the voice recognition method in the embodiment of the invention, the user voice command may be quickly and accurately recognized by the mobile terminal at any network state.

An independent learning based P2P network flow identification method is an efficient and accurate P2P flow identification method, and a deep packet inspection (DPI) method and a deep flow inspection (DFI) method as well as P2P flow identified by the DPI are used for verifying machine learning based DFI identification results to achieve automatic learning. The problems that DFI cannot be adjusted automatically and the identification rate is low are solved. According to the P2P network flow identification method with the independent learning capacity, the P2P flow is extracted through a Net Filter technology, the DPI technology is used for identification, recognized P2P flow characteristics are added in an internet protocol (IP) address list, and the machine learning based DFI identification results are verified, so that the whole identification process forms a closed-loop system.

The invention provides an image character recognition method and system based on deep learning and a medium. The method comprises the steps that the source category of an image is judged; extracting an image target area through a convolutional neural network and classifying the target area; carrying out orientation correction on the image of the target area, rotating the image to a forward orientation, and calculating an inclination angle of the image through line segment detection and a frequency domain signal analysis method; calculating a feature map of the image by using a target detectionalgorithm and a deep convolutional network, and carrying out target segmentation on the text line to carry out character recognition; according to a CRNN algorithm, combining the deep convolutional network with a bidirectional cyclic network, and carrying out end-to-end network training; and obtaining the position of the character in the picture and the model recognition content through training,and extracting character information. By adopting the computer vision and character recognition technology, the recognition problem of bill cards and table document data in the intelligent auditing process in the financial field is solved.

The invention provides a motion data monitoring method and system. A first sensor and a second sensor are used for acquiring motion data, data pictorialization and 2D image 3D synthesis are used, hitting power is measured indirectly through a motion sensor, and learning, training, sparring, feature extraction and strength and weakness countermeasures are supported, so that automatic user identification, automatic motion identification, strong event identification, weak event identification, automatic judgment and formation of automatic match odds are achieved. Besides, the functions of roll calling, sign-up, notification, positioning, alarming and the like can be achieved. The system comprises the sensors, a micro base station, an intelligent mobile phone APP, a PC, a cloud center and other hardware, cloud center software and application software.

The invention discloses a mutual conversion method of visible light and near-infrared human face images. The method comprises the steps of converting a near-infrared human face image or a visible light human face photo into an initial visible light human face photo or an initial near-infrared human face image by use of a method based on sparse learning, and converting the initial visible light human face photo or the initial near-infrared human face image into a high-definition detail photo of the visible light human face photo or the near-infrared human face image by use of a method based on multi-characteristic selection. According to the mutual conversion method, heterogeneous human face images are fitted in a stratified manner by use of a method based on sparse regularization and the visible light human face photo is generated from the near-infrared human face image, and therefore, the detail information of the synthetic photo is increased and the problem of heterogeneous human face recognition is solved.

The invention relates to a vehicle type identification method based on a multi-feature fusion neural network and a processing terminal. The method comprises the following steps: 1, training a preset neural network by using a training algorithm to obtain parameters of the neural network so as to determine the trained neural network; 2, obtaining an original image including vehicle type feature, preprocessing that original image to obtain a first image consistent with a preset pixel size, and extracting local features from the original image by adopting an object detection algorithm to obtain asecond image including the local features; 3, taking that first image as the data layer of the network, inputting the second image as the rois of the region of interest of the network to the trained neural network for feature extraction, obtaining fusion feature, classifying the fusion features by using a classification algorithm, and obtaining a classification result of a vehicle type and a probability of a corresponding vehicle type. The invention can effectively solve the problem of identifying similar vehicle types and improve the accuracy of vehicle type identification.

The invention discloses a self-adaptive stair-climbing control system and a self-adaptive stair-climbing control method. The self-adaptive stair-climbing control method comprises the steps of: establishing an exoskeleton or biped robot model according to lengths and overall weight of joints and connecting rods; acquiring obstacle distance information, and detecting whether an obstacle exists in the front is detected by means of an obstacle detection module; if so, detecting size information of the obstacle, and judging whether the obstacle is an obstacle that can be crossed by means of an obstacle classification module; if the obstacle is an obstacle that cannot be crossed, controlling a robot to move into a safety range through planning a motion trajectory by means of a safety judgment module; and if the obstacle is an obstacle that can be crossed, completing the crossing of the obstacle by means of the stair-climbing control module. The self-adaptive stair-climbing control system andthe self-adaptive stair-climbing control method solve the problem of identifying environments with unfixed height of stairs, improve the adaptability of the robot to the external environment, realizethe anthropomorphic gait of the robot, improve the overall intelligence of a machine, and have high transportability and stability.

A document desensitization system based on big data is used to desensitize Word, Excel, PPT, TXT, PDF, XML format documents desensitization processing, the system is mainly composed of system management module, data source management module, sensitive data discovery module, desensitization task management module, desensitization configuration management module, desensitization verification module,multi-level management module, security audit module seven modules. By means of natural language processing and semantic analysis, the invention solves the identification problem of sensitive data ina document, and the identification accuracy is high. The invention provides a method for solving static desensitization and dynamic desensitization of unstructured data such as documents, which ensures the safety of sharing and exchanging documents under the environment of big data. By analyzing the document, identifying the sensitive data in the document and desensitizing the document, the invention ensures that the original format of the document is not destroyed, and effectively solves the difficulty of desensitizing the document.

The invention relates to a computer vision technology, and discloses a television station logo identification system based on deep learning to improve the ability of station logo identification. The system comprises a sample collection module used for collecting station logo samples, a sample screening module used for screening samples to remove inappropriate samples, a station logo segmentation module used for segmenting a station logo from a background, a sample synthesis module used for artificially synthesizing different samples, a model training module used for training a station logo identification model based on the collected samples and the artificially synthesized samples, and a station logo identification module used for identifying station logos based on the identification model. The system of the invention is suitable for identifying normal station logos, rebroadcast station logos and extremely similar station logos.