904results about How to "The recognition effect is accurate" patented technology

The invention provides a method for detecting lane lines based on grayscale estimation and cascade Hough transform, belonging to the technical field of image processing, and relating to image segmentation, image line geometrical characteristics detection and intelligent vehicle navigation simultaneously, which is mainly applicable to an auxiliary safety drive system. The method comprises the following steps: carrying out grayscale estimation on sensitive areas of an original image I of the collected vehicle road conditions ahead; dividing the sensitive areas into a vehicle shadow region, a roadway non-mark region and a roadway mark region (including the area of the car body of the vehicles ahead); adopting a mathematical morphology method to obtain a boundary image from the roadway mark region in the areas divided by the grayscale estimation; then carrying out Hough transform on the obtained boundary image of the roadway mark region to extract line image characteristics therein; and finally realizing the detection of the lane lines by searching end points of roadway mark boundary. Compared with the congeneric method, the method in the invention has stronger stability and wider application occasions and the like.

The invention relates to a hierarchical classification method for internet flow, comprising the following steps: firstly, coarse grained classification is conducted as follows: a flow statistical property method based on machine learning is adopted for quickly classifying network flow, the network flow is divided into applied categories with different characteristics, if the flow distinguished by the coarse grained classification needs to be used, the flow can be output directly, and if the flow needs further classification, next step is conducted; and secondly, fine grained classification is conducted as follows: classification is further conducted on the applied categories distinguished by the coarse grained classification, the network flow can be intelligently, accurately and effectively identified in real time, good integrity and expandability are realized, and the requirements on the internet flow classification with different application targets and classification grain size can be met.

The invention provides a method for constructing a dense small target detection model, the model and the detection method. Based on the information fusion of the target midpoint context, through cutting the picture with high resolution, the method avoids the picture sampled under the input network from losing too much image information, and affects the network feature extraction. The residual pyramid feature extraction network is used to fuse the features of different scales, which improves the detection accuracy of the network for different size targets, especially for small targets. A RoIAlign layer instead of RoIPooling layer is used to solve the position deviation of candidate frames caused by feature mismatch of candidate regions. Because the small target features are easily lost in the network transmission, the center point context features are fused with the original RoI features, so that the network can make full use of the target context information, ensure the network runningspeed, more accurately locate and identify the dense small target, and improve the network performance.

The invention discloses a breakdown intelligent classification and positioning method of an electric transmission line. The technical scheme of the breakdown intelligent classification and positioning method of the electric transmission line breakdown is that the advantages of three kinds of technologies of a support vector machine (SVM), self-adaptation nerve fuzzy inferences (SVM) and radial based function (RBF) neural networks are concentrated. The breakdown classifiers and positioners of the SVM, the SVM and the RBF neural networks are designed. Positioning errors, classification accuracy and model operation time are used as evaluation indicators. According to the standard that accuracy is preferred and efficiency is taken into account, intelligent selection of an optimal classifier and an optimal positioner is achieved under different breakdown conditions, and optimal breakdown classification and positioning effect is achieved. Meanwhile breakdown serious extent and repair indicators are designed to evaluate breakdown injury extent and breakdown repair difficulty. The breakdown intelligent classification and positioning method of the electric transmission line effectively improves power supply reliability, reduces outage cost, and meanwhile greatly reduces workload of maintenance personnel and improves working efficiency.

The invention discloses a graphic pattern text detection method based on deep learning. The method includes firstly, training a depth convolution self-encoding network by combining graphic pattern text samples, than adopting marked samples, and classifying through a sparse dictionary; extracting graphic pattern texts from a sample library, rotating, shifting and transmitting, and combining the graphic pattern texts with pure background graphics; adopting a combined sample seat, establishing a depth convolution self-encoding network, and learning characteristic templates in layered training and entire optimizing manners; performing characteristic extraction on the characteristic templates acquired by deep network learning according to the acquired marked samples; sampling the extracted characteristics in the size of the original graphics, adopting single blocks as identifying units, and training the sparse dictionary and a classifier; after training, performing multi resolution decomposition on the graphics to be processed, utilizing the characteristic templates to extract characteristics, and utilizing the sparse dictionary to acquire results in a classified manner.

The invention discloses a gait segmentation and gait recognition integrated method based on deep learning. According to the method, human figure contour segmentation is performed on multiple gait images in a segment of gait video by utilizing a multichannel neural network segmentation model so that human figure contour segments of multiple gait images in the segment of gait video are acquired; and then identity recognition is performed on the acquired human figure contours through a classification convolution neural network model, and an identity recognition result is outputted. The method has extremely high robustness on scene changes, dressing changes, the angles of the image video and the walking state, and the method is especially suitable for solving gait recognition under the dynamic background and can achieve extremely high recognition precision in the actual gait recognition; and an integrated framework of segmentation and recognition is adopted so that the method also has extremely high recognition speed and is suitable for real-time gait recognition under the actual monitoring.

The invention discloses a rapid image-recognizing method of a power line profile. The method comprises the following steps: (1) conversing a shot color image into a grey image by brightness calculation and extracting the border profile of the image; (2) adopting an improved probability Hough transformation method to make self-adaptive adjustment of a line segment length and a statistics number obtained from the image in a known region; (3) adopting fitting of a line and a direction template for the two-dimensional fitting of a direction pixel template and a distance pixel template with linear characteristics and realizing linear fitting by deriving sumi rho(ri); and (4) calibrating and outputting power line pixels by storing and displaying the finally detected and calibrated linear pixels in a polar coordinate mode or a rectangular coordinate mode. Aiming at the concrete requirements of helicopters or robots for inspecting transmission lines, the invention precisely extracts the transmission lines in real time, provides a solid foundation for the subsequent zoom-in shooting of a camera and the defect detection of parts hereafter and has favorable theoretical and application prospects.

The present invention relates to a binocular vision-based workpiece grabbing method. The method includes following steps that: (1) a binocular vision system is calibrated; (2) a left camera and a right camera simultaneously collect images and correct the images; (3) a template matching algorithm is adopted to identify and position a workpiece in the left image and right image; (4) the three-dimensional attitude of the workpiece is calculated according to the identification and positioning information of the workpiece of the images of the left camera and the right camera; and (5) an industrial robot automatically grabs the workpiece according to the three-dimensional attitude information. The binocular vision-based workpiece grabbing method of the invention has the advantages of high real-time performance, high applicability, accurate grabbing, high working efficiency and the like.

The invention discloses a crane barrier monitoring and prewarning method and system based on binocular vision. The method includes the steps that a binocular vision model is established, and relevant parameters are initialized; images are acquired in real time and preprocessed; target edges are detected, and the characteristics of the images are extracted; the left images and the right images are registered, and a common area is segmented; iteration is repeatedly conducted, and foregrounds and backgrounds are distinguished; three-dimensional information is rebuilt, and the foregrounds and barriers are recognized; motion parameters are resolved, and prewarning information is analyzed and output. According to the crane barrier monitoring and prewarning method and system, a binocular vision system is adopted, the working environment of a crane is monitored in real time, the problem of preventing collision of the crane is effectively solved, and the system has the advantages of being high in efficiency, moderate in accuracy, simple in structure, low in cost, and very suitable for online, non-contact product detection and quality control on a manufacturing site.

The invention discloses a charging pile and a recognition method thereof, as well as an intelligent mobile robot. The charging pile comprises a pile body with an electricity supply part and a mark, wherein the mark comprises at least two mark elements with different reflex reflection coefficients and which are adjacently arrayed on one plane of the pile body in the same direction; and the mark issimple in structure and small in size, and the manufacturing cost is reduced. The recognition method of the charging pile comprises the following steps of obtaining scanning data of a laser radar by the intelligent mobile robot with the laser radar; then according to reflective strength and measuring distance, determining a first mark element and a second mark element with different reflex reflection coefficients; and finally, according to the reflective strength, the measuring distance and the angle value, determining the pose of the intelligent mobile robot relative to the charging pile, sothat after the position of the electricity supply part is determined by the intelligent mobile robot, the intelligent mobile robot is in butt joint with the electricity supply part for charging, so that accurate recognizing and accurate positioning are realized.

The invention discloses an identification method of fine-grained attributes of pedestrians under complex scenes. A classification model is adopted for fine-grained attribute identification on sub-components of a detected pedestrian; association analysis is respectively carried out on identified attributes and a pedestrian gender, and attributes with high correlation are selected for multi-task learning; then convolutional neural network (CNN) models constructed by multi-task learning are trained, and results of the convolutional neural network models with a highest identification rate are selected for the multiple attributes to use the same as a final result; and finally, a gender attribute of the pedestrian is judged according to a customized decision function. The method can realize overall-to-local detection of the pedestrians in the complex scenes, realize more accurate detection and identification of the attributes of pedestrian sub-components, and avoid interference of backgrounds and other information, also solves, at the same time, the problem that detection correctness rates of models on small objects are low, and has higher information accuracy.

The invention provides a treatment method for a surface target of an unmanned ship based on a laser imaging radar.The method based on the unmanned ship of the laser image radar comprises the following steps: S1, generating a three-dimensional cloud point image on the water surface around the unmanned ship by the laser imaging radar, the three-dimensional cloud point image comprises a target cloud point and a non-target cloud point;conducting dimension reduction treatment to the three-dimensional cloud point image, projecting the three-dimensional cloud point image to a two-dimensional XY-grid plane, counting the position information and height information of each grid, S2, cutting the target cloud point and non-target point cloud, S3,clustering the target point cloud obtained after being cut, extracting the position information of each target, forming the target sample set, extracting multi-dimensional eigenvector collected by the target sample; S4, training the target sample set,obtaining the obtained identifying function, and identifying the target point cloud by the identifying function. The treatment method provided in the invention can detect and identify the target of the water surface around the unmanned ship accurately.

The invention provides a mobile robot and a positioning method thereof. The method includes the following steps that: the voice information of a user is acquired, the direction of the user is positioned according to the voice information; the mobile robot is rotated so as to make an image positioning unit face the direction; an image sequence is acquired, and the feature portion of the user in the image sequence is identified; and according to the position of the feature portion, the position of the user is calibrated, and the distance of the user is positioned. According to the mobile robot and the positioning method thereof of the invention, image acquisition is guided through voice positioning, the feature portion of the user in the image sequence is identified, according to the position of the feature portion, the position of the user is calibrated, and the distance of the user is positioned, and therefore, accurate positioning of the user who is adopted as a sound source can be realized; and the summoning or tracking commands of the user are further identified, and the mobile robot is driven to a position in front of the user after the user is accurately positioned.

A pedestrian re-identification method, system, electronic device and computer-readable storage medium are disclosed. The method includes: acquiring a video set and determining a target pedestrian image; Extracting target features of the target pedestrian image, and inputting the target features into the NSN to extract an image to be recognized of each frame of the image in the video set; Generating an optical flow map of an image to be recognized of adjacent frames by using an MN network; All the images to be identified and all the optical flow images are input into the LSTM network to obtaina fusion picture fusing multi-frame image information. The global feature map and partial attention map of the fusion image are extracted from the trained human feature extraction network, and each part of the attention map and the global feature map are fused into a fused partial attention feature map respectively. The fusion feature vectors of each part of the attention fusion feature map are formed by the global average pool, and all the fusion feature vectors are connected to the global feature vectors, which improves the recognition accuracy of pedestrian re-recognition.

The invention discloses an intelligent fruit picking robot based on machine vision. The robot comprises a control system and a walking mechanism, a picking mechanical arm, a visual system and a powersystem which are connected with the control system respectively, the control system, the power system and the walking mechanism are fixedly connected, and the picking mechanical arm is a four degreesof freedom joint mechanical arm and is fixedly connected with the walking mechanism through a base; a visual sensor of the visual system is fixedly arranged above the wrist of the picking mechanical arm, a visual control module of the visual system is integrated with the control system and is connected with the visual sensor through signals, and the visual control module is used for conducting target positioning and identifying on images of the visual sensor based on a full-convolution network image processing technology of image partition; the visual control module transmits target positioning information to the control system, and the control system controls the walking mechanism to move and / or the mechanical arm to pick fruits. The intelligent fruit picking robot based on the machine vision has the advantages of compact structure, good stability, high intelligent degree, quick positioning, high positioning precision, and high environmental adaptability.

The invention relates to a method for predicting city traffic accidents based on time-space distribution characteristics. The method comprises: first, in combination of the case information and the space information, creating a case space database and performing pretreatment to the data; then, based on surface area statistics, analyzing the traffic accidents' time-space distribution characteristics; using the global and local self-correlation method to realize the analyzing of the aggregate state; based on the case happening point data, analyzing the traffic accidents' time-space distribution characteristics; through the hierarchical clustering analysis, expressing the distribution rule of the cases hierarchically; through the nuclear density estimation method, expressing the continuous changes and accurate gathering center of the traffic accidents' happening distribution; and finally, utilizing the BP neural network prediction algorithm, using the time-space distribution characteristics of the already happened cases to predict the time-space distribution areas of traffic accidents in the future. According to the invention, in combination with the time-space distribution and through the utilization of big date excavation BP neural network prediction algorithm and the time-space distribution characteristics of the already happened cases to predict the time-space distribution areas of traffic accidents in the future, it is possible to increase the precision, the timeliness and reduce the manual cost.

The present invention discloses a system and method of library service based on a robot. A RFID electronic tag and a radio frequency inductor are employed to realize construction of a library indoor map and the position location of a robot, and the robot realizes active navigation according to the positions of the library map and the robot itself to facilitate allowing the robot to directly drive to a shelf where a book is located. The ingenious design of a mechanical structure for book access is employed to reduce the cost and facilitate control. bar codes are pasted at the spine of a book and a bar-code identifier is combined to facilitate realization of book location and allow the identification effect to be more accurate. The robot performs data excavation according to the book information base and reader borrowing recording base of the library to push books which accord with the readers' interest and provide better service experience for the readers. The robot can provide service such as beverage sale, coffee ordering, free drinking water except for basic book borrowing service so as to allow the readers to enjoy a more comfortable reading environment.

The invention discloses a method for voice recognition and electronic equipment. The method for voice recognition is applied to the electronic equipment, and the method includes: receiving first voice data input by a user; based on identity information of the user, obtaining first high-frequency vocabulary database corresponding to the identity information of the user; and utilizing the first high-frequency vocabulary database to recognize the first voice data. Corresponding voice recognition is performed on the voice data input by the user through the high-frequency vocabulary database corresponding the user, thereby solving the technical problem in the prior art that since all voice data need to be compared and recognized in a standard vocabulary database, the efficiency of voice recognition is low, and achieving the technical effect of improving the efficiency of voice recognition.

The invention discloses an enhanced robot workbench based on multiple depth cameras. The enhanced robot workbench comprises a workbench, a mechanical hand arranged on the workbench, a plurality of depth cameras arranged around the workbench, and a control processor. According to the enhanced robot workbench, the visual perception of the robot workbench is enhanced through the multiple depth cameras, and a target locating method is more convenient and more rapid through adopting an automatic target locating method of the depth cameras and the robot; the three-dimensional coordinate of a target object can be recognized and located more accurately through utilizing the multiple depth cameras in the situation that the object is shielded; and during target object recognition, the image recognition efficiency is greatly improved through narrowing the image matching area.

The invention relates to a fake certification method for mobile phone electronic tickets and a system therefore. The system comprises a mobile terminal (1), a computer terminal (2), a communication network (3) and a server (4), wherein the server is provided with an internet verification subsystem, a SMS verification subsystem and an anti-counterfeiting code list generation subsystem under software support. The invention can provide various verification tools including mobile phone, computer, anti-counterfeiting code list and mobile phone two-dimension code special reader from the website of electronic ticket service to trade companies and members, and a method thereof, therefore, trade companies can quickly and accurately verify the reality, falsity and effectiveness of mobile phone electronic tickets without using expensive machine recognizers, to reduce the fake certification cost on electronic tickets significantly, thereby exciting the enthusiasm and activeness of trade companies to enter into website to publish electronic ticket information, and developing the websites of various electronic ticket services effectively and quickly.

The invention discloses a fault plane identification method based on a full convolution neural network, which comprises acquiring seismic amplitude fault data, constructing a full convolution neural network model, training a full convolution neural network training model, and identifying seismic amplitude fault data. The present invention extends from fault classification identification at the level of seismic amplitude images to classification at the level of pixels, using full convolution neural network to segment 3D seismic amplitude data can not only realize fast fault recognition, but also avoid the loss of spatial information by using full convolution instead of full connection layer, so as to obtain more accurate fault segmentation and recognition.

A capacitive touch screen and a single layer wiring electrode array. The single layer wiring electrode array includes a capacitive region and a wiring region located on one plane; the wires of the wiring region are in a zigzag or a wave pattern. A capacitive touch screen comprising a single layer wiring electrode array, the capacitive touch screen including a substrate; a single layer-wiring electrode array, disposed on the substrate, the single layer wiring electrode array includes a capacitive region and a wiring region located on one plane, the wires of the wiring region are in a zigzag or a wave pattern; the wires of the wiring region connected to control ports corresponding to at least one or more integrated circuit chips respectively. The single layer wiring electrode array and capacitive touch screen makes the fabrication cost lower, the display effect better.

The invention provides a remote sensing target detection method based on sparse guidance and significant drive. The method comprises the steps of (1) dividing an input remote sensing image into sub blocks, extracting global color feature clusters to form a global dictionary and extracting the color feature clusters of the image edge sub blocks to form a background dictionary, (2) carrying out sparse expression on all input image sub blocks by using the global dictionary and the background dictionary and obtaining global and background sparse expression coefficients, (3) clustering the sparse expression coefficients obtained in the step (2) to generate global and background significant maps, (4) carrying out smooth denoising on the global and background significant maps obtained in the step (3) and then obtaining a final significant map by using Bayesian fusion, and obtaining a significant target area, (5) extracting a color feature and a texture feature for the significant target area detected in the step (4) and a collected training sample, and carrying out sparse expression by using a maximum value restraint sparse coding model, and (6) carrying out target type identification on the significant target area by using a sparse expression coefficient obtained in the step (5). According to the method, an interested remote sensing target can be accurately and rapidly detected and identified in a complex background, and the effect is outstanding.

The invention discloses a universal serial bus (USB) device and a method for recognition of a host operating system. The method comprises the following steps: waiting for and receiving a USB command coming from a host by the USB device, judging whether the received USB command is a command for obtaining a deployment descriptor or not, if yes, confirming the host operating system according to values of a first identification bit and a second identification bit and values of length bytes in the command for obtaining the deployment descriptor, after the host operating system is confirmed, utilizing a corresponding communication protocol to carry out data communication with the host according to the host operating system, shielding files which cannot be operated under the host operating system, enabling the host to recognize and operate the USB device accurately, and enabling the USB device and the application of the USB device to be optimized.

The invention relates to an emotion recognizing and tracking method based on video information. The emotion recognizing and tracking method based on video information comprises the steps that 1, an image is obtained, and three-dimensional head modeling is conducted on the obtained image; 2, image fusion is conducted by means of a generated three-dimensional head model, and continuous emotion expressions and emotion expressions having nothing to do with people are formed; 3, a training sample is constructed by means of a generated fused image, the three-dimensional head model and the emotion values of the three-dimensional head model; 4, an emotion recognition model is trained by means of the constructed training sample; 5, a series of pretreatment operations are conducted before testing is executed if necessary, three-dimensional head key point tracking and emotion recognizing are conducted by means of the trained emotion recognition model till all emotion recognition tasks are completed. The emotion recognizing and tracking method based on video information is suitable for recognition of discrete emotions as well as for emotion recognition under the continuous emotion space expression condition, can be used for tabletop scenes and movable interactive scenes, is not limited by visual information acquisition equipment and can improve the natural man-machine interaction quality.

The invention discloses a recognizing and classifying method for broadband radar interference signals. The method comprises the following steps of S1, obtaining radar interference signals, processingthe signals, and obtaining digital video signals; S2, performing short-time Fourier conversion on the digital video signals, and obtaining a time frequency matrix; S3, according to a ranking statistics constant false-alarm method, processing the time frequency matrix, and obtaining a constant false-alarm detection matrix; S4, calculating parameters of interference signals in the constant false-alarm detection matrix, and integrating the parameters of the interference signals into a signal parameter matrix, and obtaining an interference signal detection result; S5, according to the interferencesignal detection result, calculating interference signal characteristics; S6, according to the interference signal characteristics, recognizing and classifying the interference signals. According tothe interference-based time-frequency signal characteristics, nine broadband radar interferences are recognized and classified, and the method has the capability of multi-interference (non-overlappedon frequency) classifying and recognizing.

The invention discloses a vehicle retrograde motion intelligent detection method based on tracking track analysis. The method comprises the following steps: detecting a target vehicle in a video imageby utilizing a primary classifier and a secondary classifier; extracting a target vehicle area, distributing a nuclear correlation filtering tracker for the target vehicle; wherein each target vehicle is matched with one nuclear correlation filtering tracker, the nuclear correlation filtering tracker is used for obtaining a tracking area of the target vehicle, then the movement track growth direction of the target vehicle is obtained according to the tracking area and compared with the initially marked retrograde movement direction, and if the movement track growth direction is the same as the initially marked retrograde movement direction, the target vehicle retrograde movement occurs; And if not, the target vehicle does not run reversely. By means of the mode, real-time detection of vehicle retrograde running is achieved, and the problems existing in manual detection are solved; And meanwhile, the reliability of vehicle detection and recognition is improved by using the cascade classifier, and simultaneous tracking of multiple vehicle targets is realized.

The invention discloses a method for identifying a low-voltage distribution network topology and a line impedance in a station area, and belongs to the technical field of a low-voltage distribution network. The method comprises the steps of adding an edge computing terminal and a plurality of electrical measurement devices to the low-voltage distribution network in the station area firstly; when the low-voltage distribution network topology of the station area is identified each time, determining the affiliation relation of the station area of the measurement devices and a phase type of a communication access phase according to a power line carrier communication relation between the measurement device and the edge computing terminal, enabling each measurement device and the edge computingterminal to sample voltage and current waveforms after time synchronization and updating the voltage and current waveforms to the edge computing terminal, and enabling the edge computing terminal to identify the same bus-bar measurement device and an upper-level measurement device of the bus-bar according to the waveform data; and after the topology is identified multiple times, obtaining a finaltopology identification result and performing impedance calculation. According to the method for identifying the low-voltage distribution network topology and the line impedance in the station area disclosed by the invention, the accuracy rate for identifying the low-voltage distribution network topology and the line impedance is high, the speed is rapid, the information collection capability of the intelligent equipment of the low-voltage distribution network is fully exerted, the equipment cost is low and the method has no influence on power quality.

The invention discloses an openCV(open source computer vision library)-based video target tracking algorithm. The openCV-based video target tracking algorithm is characterized in that a selected template is matched with a video frame; a subgraph position which is most similar to a template image in the video frame is found out through calculation of correlative coefficients; updating of the template is determined according to a predicted position of a kalman filter and the correlative coefficient values; and the specific algorithm includes template matching, position prediction and template updating. Compared with the prior art, the openCV-based video target tracking algorithm has the advantages that recognition and tracking of a target are not affected by environment change; the target object is accurately recognized and tracked in time; and the updating of the tracked object is available, so that the template can be dynamically updated during the system tracking, and the tracking is more accurate under the condition of environment and object change. In addition, a plurality of parameters are used as tracking evidences, so that the tracking is more reliable; and under the condition of target object moving, continuous background change and shadow influence, the tracking object cannot be lost..

The invention discloses a coin identifying and sorting device and method. The coin identifying and sorting device comprises an obliquely-arranged coin channel; the front end of the coin channel is provided with a slit; the coin channel is provided with a coin diameter identification system, an electromagnetic identification system, a weight identification system and a coin sorting system; the coin diameter identification system comprises a photoelectric bi-transmission sensor arranged on the side wall; the electromagnetic identification system comprises a coil and an LC oscillating circuit which are arranged on the side wall; the weight identification system comprises a detection plate which is arranged at the bottom of the coin channel and descends under the action of gravity; the coin sorting system comprises a coin outlet arranged at the back end of the coin channel; the coin outlet comprises a counterfeit coin outlet and real coin outlets of different types; a guide plate for closing corresponding coin outlet is hinged to every coin outlet; every coin outlet is provided with an electromagnet for controlling a corresponding guide plate to rotate. The coin identifying and sorting device and method identifies the authenticity and type of coins by detecting a plurality of features of the coins and can accurately identify and sort the coins.