174results about How to "Recognition stability" patented technology

When authentication data of a person O to be authenticated is registered as dictionary data, this authentication data of the person to be authenticated is acquired and collated with the registered dictionary data. In accordance with the collation result, the dictionary data is updated.

A method for quickly identifying lanes-change condition of object vehicle includes utilizing machine vision and radar to collect horizontal and longitudinal distances, relative speed and the same lane possibility between object vehicle and local vehicle, making comparison and quantization on said data collected each time, making compensation and tracking prediction on relevant data, calculating confidence of each vehicle by using two layers of neural network training to obtain object vehicle weight and setting object vehicle to be dangerous object once calculated confidence coefficient is over confidence threshold.

The invention discloses a bone age mark identification assessment method and system based on deep learning and image omics. The bone age mark identification method includes the steps: performing preprocessing of window adjusting, alignment and standardization on the wrist bone image; using a bounding box to mark the bone age characteristic areas and mark the coordinates, wherein the bone age characteristic areas include a metacarphphalangeal group and a brachidium group according with a TW3 method; according to the requirement, performing augmentation processing, and inputting the wrist bone image data to a convolutional neural network of the area based on ResNet-101 to perform multi-task (positioning, classification and assessment) training at the same time; and based on the bone age characteristic areas, combining with the clinic information (demographic characteristics and inspection reports) to further train and improve the bone age assessment speed and accuracy. The bone age markidentification assessment method and system based on deep learning and image omics firstly utilize a small number of marked samples to perform preliminary training on the bone age model, and utilize the model with relatively higher positioning detection accuracy to automatically mark a large number of samples so as to realize automatic positioning, classification and bone age assessment of the bone age characteristic areas.

A lane marker recognition method comprises the steps of inputting an image including a lane marker, extracting luminance change points in the image, and estimating the position of the lane marker using the extracted edge points. The edge extracting step includes calculating an angle of orientation of each of the edges and the lane marker position estimating step includes extracting edge points that are oriented toward a vanishing point of a road, from among the edge points extracted in the edge extracting step.

The invention belongs to the technical field of autonomous running and visual perception, particularly relates to a multi-camera visual perception system for a UGV (Unmanned Ground Vehicle) and aims to solve the problems of single perception algorithm, poor real-time performance, poor system expandability, inconvenience in debugging, limitation to the view field control and the like of an image processing system of the existing UGV and autonomous mobile robot. The system comprises a plurality of vehicle-mounted image processing industrial personal computers and one or a plurality of debugging monitoring computers, wherein each vehicle-mounted image processing industrial personal computer is connected with three cameras; one or a plurality of debugging monitoring computers are used for controlling the vehicle-mounted image processing industrial personal computers by a wireless router; and a control mode can be a redundancy control mode and also can be a respective control mode. In the system, an optimized visual perception method on the basis of multiple threads is adopted; and threads comprise a multi-camera video acquisition thread, a lane line perception thread, a barrier, cone bucket and signal lamp perception thread and the like. Compared with a conventional single-thread visual perception method, the optimized visual perception method on the basis of the multiple threads has higher real-time performance, stability and expandability.

A community access control system based on face recognition is characterized by comprising a face information database, an image acquisition module, a face detection module and a face recognition module. The face information database is used for storing face images and feature data of permanent residents in a community. The image acquisition module is used for acquiring face video information of the residents and obtaining face pictures from the face video information. The face detection module conducts preprocessing on the face pictures to obtain a face image sequence. The face recognition module conducts facial feature extraction on the face image sequence of the face detection module, obtains face feature data, matches the face feature data with the feature data in a face feature database, and outputs matching results. According to the community access control system based on the face recognition, a face recognition software system is divided into three mutually independent modules, namely the face detection module, the face recognition module and the face information database, so that the face recognition software system is stable in recognition and high in running speed. Meanwhile, through preprocessing of collected image information, the overall community access control system can overcome the influence of illumination variation, and therefore quite high recognition can be realized no matter in what kind of light environment.

A method and a device for recognizing a pedestrian and a vehicle supporting the same are provided. The method includes collecting, by a controller, a far-infrared image using a far-infrared imaging device and detecting a pedestrian candidate group from the far-infrared image. In addition, the method includes extracting, by the controller, pedestrian features based on previously normalized pedestrian database (DB) learning and comparing the pedestrian features with the pedestrian DB learning results to determine similarity. The controller is configured to perform pedestrian recognition based on the comparison result.

A reader antenna includes dipole antennas, the dipole antennas including: a horizontal dipole antenna having the characteristic of horizontal polarization; and a dipole antenna perpendicular to the horizontal dipole antenna and including a vertical dipole antenna having the characteristic of vertical polarization, wherein the horizontal dipole antenna and the vertical dipole antenna may both have the characteristic of circular polarization. The reader antenna can be easily stored in an electronic shelf because it is realized in an ultrathin structure using a single-layered dielectric substrate.

An electronic device is disclosed. The electronic device comprises: multiple microphones for receiving audio signals generated by multiple sound sources; a communication unit for communicating with a voice recognition server; and a processor for determining the direction in which each of the multiple sound sources is located with reference to the electronic device, on the basis of the multiple audio signals received through the multiple microphones, determining at least one target sound source among the multiple sound sources on the basis of the duration of the determined direction of each of the sound sources, and controlling the communication unit such that the communication unit transmits, to the voice recognition server, an audio signal of a target sound source from which a predetermined voice is generated among the at least one target sound source.

The invention relates to a method for automatically detecting a pre-defined image pattern in an original picture, wherein pixel data from said original picture are looked through by means of a processing step, including at least one transform, to find said pre-defined image pattern, wherein according to the invention said processing is split up into at least two stages, wherein a first stage with a coarse processing is to detect locations in the original picture imposing an increased likelihood that the pre-defined image pattern, can be found there, and wherein a second stage with a refined processing is applied to the locations to identify the pre-defined image pattern.

The invention discloses a technology for identifying and tracking a human eye. The technology comprises the following steps: a. initiating an iris spatial position acquisition module; b. under visible lights, a single camera acquiring images of a human face or human eye; c. applying a human face or human eye calibration method to process image information acquired under visible lights, obtaining image coordinates of feature points of the human face or human eye that includes an iris center; d. using the single camera to reconstruct the iris center spatial position; e. obtaining a human eye iris center spatial position which including spatial position information of the human eye with respect to a display.

The invention discloses a pumping unit indicator diagram identification method based on an ART2. By collecting load and suspension center displacement signals of a pumping unit, a closed curve formed along with time changes in a pumping period is drawn with polished rod displacement signals serving as an x-coordinate and loading signals serving as a y-coordinate, and a pumping unit indicator diagram sample is obtained. If the category of the pumping unit indicator diagram sample is known, the pumping unit indicator diagram sample is identified through a mode with a supervision module. If the sample category of the pumping unit indicator diagram sample is unknown in advance, the pumping unit indicator diagram sample is identified through a mode without the supervision module. The pumping unit indicator diagram sample is identified through an ART2 neural network, according to an identification result, fault diagnosis and underground work condition judgment are performed on the pumping unit, and a judgment result is displayed. The pumping unit indicator diagram identification method based on the ART2 has the advantages of performing fault diagnosis and underground work condition judgment on the pumping unit, identifying the indicator diagram stably, efficiently and accurately, and the like.

The invention relates to a radiation source identification method and device based on bispectrum three-dimensional image texture features. The method comprises the following steps: acquiring sample data of different radiation source signals and calculating bispectrum of signals to obtain bispectrum three-dimensional images; for bispectral three-dimensional images, extracting texture features by agray level co-occurrence matrix; for texture features, using a support vector machine to classify and identify the emitter. The invention avoids the problems of high characteristic dimension and lossof integral bispectrum part classification information and the like existing in adopting bispectrum alone. The steady-state signal RF fingerprint in individual identification of an emitter is extracted, sample information is fully utilized, the efficiency, application range and robustness of classifier training and recognition are effectively improved, and the advantages under different conditionssuch as noise immunity, small sample size, multi-target, and universality of different modulation modes are exhibited. The effectiveness of the classifier in the actual electromagnetic environment isverified by the actual radio signal, and the method and the device have important research values in improving the efficiency and robustness of classifier.

An apparatus and method of recognizing a position and direction of a mobile robot includes obtaining absolute coordinates at a current position of the mobile robot and relative coordinates for a moving displacement of the mobile robot. Therefore, the position and direction of the mobile robot are recognized by reflecting the relative coordinates on the absolute coordinates. Accordingly, the present invention operates odometry and RFID coordinate systems in combination with each other, thus obtaining a high sampling rate by the odometry coordinate system while restricting an error range within a predetermined level by the RFID coordinate system.

The present invention discloses a speech spectrum fusion feature-based mobile phone source identification method. According to the method, fast Fourier transformation is performed on each frame in each speech sample in a speech library corresponding to each mobile phone, so that the amplitude of the frequency value of each frequency point of first J/2+1 frequency points of each frame is obtained; the statistical average information amount features, statistical average logarithmic spectrum features and statistical average phase spectrum features of each speech sample are obtained according to the amplitudes of the frequency values and are adopted to form spectrum fusion features; the spectrum fusion features of all the speech samples corresponding to all the mobile phones are combined to form a spectrum fusion feature matrix; normalization and feature selection are performed on the spectrum fusion feature matrix, so that an optimal feature matrix can be obtained, namely, the final spectrum fusion features of each speech sample are obtained; and model training and test sample identification are performed on the final spectrum fusion features of all the speech samples. The method has the advantages of high identification accuracy, high stability and strong generalization ability.

The invention relates to a structure characteristic-based character recognition method. In order to solve defects in the prior art, the technical scheme of the invention comprises the following steps: 1, a serial number area is positioned on an RMB gray scale image, a gray scale image for the serial number is acquired, and tilt correction is carried out; 2, pretreatment is carried out on the gray scale image after tilt correction, a binarization threshold is determined according to a brightness histogram of a target image, binarization processing is carried out on the target image, and a binarization image is acquired; 3, connected component labeling is carried out on the binarization image, and a binarization image of a single character is extracted according to a labeling result; and 4, finally, according to character structure features, the binarization image of the single character is recognized.

A speech recognition apparatus, method and computer program product whereby noise is subtracted from an input speech signal by a plurality of spectral subtractions having differing rates of noise subtraction to produce plural noise-subtracted signals, at least one speech features is extracted from the noise-subtracted signals, and the extracted feature is compared with a standard speech pattern obtained beforehand to recognize the speech signal based on a result of the comparison. In addition, features can be extracted from at least one of the noise-subtracted signals and also the input speech signal for comparison with the standard speech pattern. Plural features can be combined into a single feature for the comparison.

A speech recognition apparatus, includes a reliability estimating unit configured to estimate reliability of a time-frequency segment from an input voice signal; and a reliability reflecting unit configured to reflect the reliability of the time-frequency segment to a normalized cepstrum feature vector extracted from the input speech signal and a cepstrum average vector included for each state of an HMM in decoding. Further, the speech recognition apparatus includes a cepstrum transforming unit configured to transform the cepstrum feature vector and the average vector through a discrete cosine transformation matrix and calculate a transformed cepstrum vector. Furthermore, the speech recognition apparatus includes an output probability calculating unit configured to calculate an output probability value of time-frequency segments of the input speech signal by applying the transformed cepstrum vector to the cepstrum feature vector and the average vector.

The invention discloses a point-like distribution cooperation coding sign and an identification and positioning method. The coding sign comprises at least 6 circular sign points. The sign points comprise mark points A, B, C and O. A positioning point comprises a positioning point D. Other sign points are coding points. The mark points A, B, C, and O are located at the four vertex angles of a square respectively, and the mark point A and the positioning point D are located at the same diagonal line of the square. The mark point O is located at the extension line of a connection line between themark point A and the positioning point D, and the original point of a coding sign design coordinate system is defined as the mark point O. The X axis and the Y axis of the coding sign design coordinate system are parallel to the sides of the square. Each coding point is distributed near the X axis and the Y axis, and the coding points at different positions correspond to unique coding character values. The coding character values corresponding to each coding point form a unique coding value. In the invention, a rich number of unique coding signs can be realized, and a recognition algorithm issimple and fast, and the coding sign is positioned accurately.

A sheet paper identification apparatus and method for stably identifying a sheet paper while preventing erroneous identification caused by dust adhered to an imaging section and caused by fluctuation in illumination brightness. An imaging section (7) images the interior of the imaging section (7) before a sheet paper is fed to capture image data on an image showing no sheet paper. When a sheet paper is fed into the imaging section (7), the imaging section (7) images the sheet paper and captures image data on an image showing a sheet paper. An image data analyzing section (8) creates a density histogram of the image showing no sheet paper from the image data on the image showing no sheet paper and detects dust adhered to the imaging section (7) on the basis of the density histogram. If the amount of particles exceeds a reference value, the image data analyzing section (8) sends a report about the abnormality of the imaging section (7) to a control section (2). The image data analyzing section (8) creates a density histogram of the image showing a sheet paper from the image data on the image showing a sheet paper, determines brightness of the image on the basis of the density histogram, and adjusts imaging operations of the imaging section (7) according to the determination.

The present invention comprises an inspection method for transparent or reflective inspection objects, by using a retro-reflective plate as a convex lens or concave mirror in such a way that defects can be detected in a stable fashion even if the inspection object is curved or if vibration occurs during movement of the inspection object; in which displacement of parallel light passing through or reflected in a measurement field is captured by using changes which occur in an inspection field, namely changes in the density gradient in the case of transmission inspection or in the reflection angle in the case of reflection inspection. The present invention also comprises an inspection device for defect inspection, in which a knife edge is provided, horizontally to the optical axis, on the front surface of a camera lens for collecting the transmitted light or reflected light, in such a way that three-dimensional defect images can be obtained since any light which diverges from the parallel light is blocked with consequent changes in the contrast of the light due to the density gradient in the inspection field at the camera.