254results about How to "Ensure real-time requirements" patented technology

The invention relates to an image stitching processing system, which belongs to the field of image processing, and is characterized in that an image reading unit, an overlapping region unit, a preprocessing unit, an optimal stitching line unit, a feature location unit, a feature matching unit, and a deformation propagation and image reconstruction unit are arranged in a digital processing circuit, wherein the image reading unit is used for reading two adjacent images from an image sequence; the overlapping region unit is used for acquiring an overlapping region of the two images; the preprocessing unit is used for preprocessing color of the overlapping region; the optimal stitching line unit is used for obtaining an optimal stitching line in the overlapping region; the feature location unit is used for detecting feature points in an optimal stitching line neighbourhood; the feature matching unit is used for matching the detected feature points; and the deformation propagation and image reconstruction unit is used for obtaining a deformation vector of each pixel point of each image, obtaining a gradient field of each image, and reconstructing a stitched image according to the gradient field. According to the image stitching processing system, the optimal stitching line is used as the most important image feature, so that the limitation of a shape of the overlapping region is solved, a lapped seam can be well eliminated in color and structure, and a real-time processing requirement can be met by utilizing an acceleration algorithm.

The invention discloses a stock market investment decision-making method based on network analysis and multi-model fusion. The method includes the steps that fundamental information is grabbed from a network and then network nodes and network connection are constructed; a complex social network model is constructed; an investment portfolio is selected by means of a network analysis method and then data involved in the investment portfolio are input into a multi-model fusion frame, wherein the multi-model fusion frame comprises a plurality of sub-models; all the sub-models conduct market trend prediction with different characteristics according to technical information of different characteristics grabbed in the network generate predicted values of the corresponding sub-models, the predicted values are weighted and summated to obtain a comprehensive market trend predicted value, and then corresponding investment strategies are generated according to the comprehensive market trend predicted value. According to the method, risk factors, which are ignored in general researches, of the investment portfolio are comprehensively considered from multiple angles, real-time requirements of the strategies are guaranteed through the methods such as dimension reduction data preselection of fundamentals and technical feature selection, and consequently the more reliable investment strategies are provided.

The invention introduces a fish-eye image correction method based on multistep correction, and aims at overcoming the disadvantages of image magnification and loss of edge information caused by a normal correction model in the prior art. A correction model is established based on multistep correction. A size zooming factor, based on a polynomial model, of a fish-eye image is established; the fish-eye image is corrected for the first time by utilizing the size zooming factor; interpolation is carried out on the image corrected for the first time in a neighborhood interpolation method; a division model based on a backward mapping method is established; and via the division model, the image after interpolation is corrected for the second time to obtain a secondary corrected image. The fish-eye image is corrected via the correction model, the problems including loss of edge information, image magnification, distortion and incomplete correction of the corrected images are solved, most information of a fish-eye lens is completely reserved, and the condition that the edge of the image is fuzzy is changed.

The invention describes a robotic hybrid system application frame based on a multi-core processor architecture. In a robotic system which takes an ARM (Advanced RISC Machines)/X86 multi-core processor as a controller, a multi-core parallel processing structure of the ARM/ X86 multi-core processor is used for operating the robotic hybrid system application architecture consisting of a real-time operating system, non-real-time operating systems and a system supporting frame on a whole robotic controller so as to provide improved operating system service, wherein the real-time operating system, non-real-time operating systems and the system supporting frame simultaneously operate. In the application frame, one real-time operating system independently operates in one ARM/X86 core, meanwhile, a plurality of non-real-time operating systems operate in other ARM/X86 cores, the operating systems mutually individually occupy processor resources and peripherals, robotic application programs with different real time requirements are independently operated, and the application programs can be used (drawing 1) in a uniform ROS (Robotic Operating System) application node form.

The invention discloses a method for classifying network abnormal events. The method disclosed by the invention comprises the following steps of: firstly, extracting flow characteristic parameters from network flow data; and, determining a historical time window and constructing abnormal characteristic multi-time sequences according to a detected abnormal time point, wherein the accuracy for classifying abnormal events of a backbone network can be effectively improved by sufficiently utilizing the time correlation of a flow behavior; simultaneously, different kinds of abnormal events having similar characteristics can be effectively distinguished by analyzing the similarity between the abnormal characteristic multi-time sequences; and the real-time requirement for classifying the abnormalevents of the backbone network can be ensured by utilizing an information entropy to process mass data.

The invention discloses a 3D face cartoon animation method based on region segmentation and subsection learning. The invention has the following steps of: first obtaining sparse 3D movement capturing data and 3D face surface mesh; next calibrating the control point corresponding to the movement capturing data on the 3D face gridding; separating the rigidity and non-rigidity movement in the movement capturing data, and aligning the movement capturing data and the face model in static state; clustering the acmes of the face model based on the cosine distance of the displacement vector of the movement capturing data and the manifold distance of the 3D face gridding acmes; training the segmented surface patch to be subsection radial basis function; and fusing the border movement via the advanced Voronoi-Cell arithmetic. The invention can execute the automatic regional segmentation to a single face model, and model the face non-rigidity movement via subsection radial basis function to form the face cartoon with the third dimension.

The invention relates to an embedded-based open set speaker recognition method and a system thereof, which relates to speaker recognition in the field of voice signal processing. The embedded-based open set speaker recognition method comprises the following steps: firstly, pre-processing and characteristically extracting the voice signals; secondly, taking exercises referencing the speaker voice model; and finally, making the open set mode matching. The recognition system comprises a voice signal processing unit which is respectively connected with a voice signal collecting unit, a data storage unit, a man-machine interface unit and a communication interface unit, a power management module is connected with the voice signal processing unit, and the man-machine interface unit comprises a state indicating module, a keyboard module, a display module and a startup module. The invention has the advantages of suiting different application occasions and raising the security performance of systems. Designing an embedded system performs embedded realization of the open set speaker recognition method. The system has the characteristics of high real-time performance, good recognition performance, low power consumption, low cost, high portability, high popularity, etc.

The invention discloses a talkback and video data transmitting method based on a wireless local area network, and a talkback commutator, which are realized by adopting a plurality of wireless local area network user terminals, a plurality of wireless local area network wireless access stations, a wireless local area network switch and a talkback and video management server, can provide the characteristics of high connection speed, strong instantaneity, high voice quality, and absolute forced prior communication in priority level for talkback and video data transmission of the wireless local area network user terminals under the condition of keeping the prior functions of data access and transmission of the wireless local area network on the basis of an equipment network and external line transmission system provided by the existing wireless local area network equipment, and have functions of general calling, group calling and notification. Meanwhile, during needing, the communication connection between the wireless local area network user terminals and public telephone communication network users can be realized, so that demands on talkback and video data transmission of the wireless local area network user terminals are met.

The invention is suitable for the field of vehicle detection, and provides a vehicle detection method based on deep learning, and the method comprises the following steps: S1, obtaining a plurality ofimages containing vehicles through a video stream, carrying out the manual marking of the images, and carrying out the dividing of the images into a training set and a verification set of a detectionmodel according to a proportion; S2, constructing a target detection PDN network based on the DDN; S3, performing model training on the target detection PDN network by using a training set, and selecting a model with an optimal performance of a verification set; And S4, on the basis of the optimal model, performing prediction on the GPU server to realize target detection on the video stream. Thetechnical problems that in the prior art, in actual target detection, a video is usually processed, an algorithm is needed to carry out target detection in real time, a small part of researchers' methods meet the requirements on speed, however, with the cost of sacrificing precision, the precision is low, an application cannot fall to the ground, and a large amount of missed detection and false detection cannot be accepted are solved.

This invention provides a low power real-time task scheduling method oriented embedded system for modeling, solution and optimization of real-time task scheduling for low power. It achieves low power consumption by using the technology and Real-time dynamic FM_PM Web tablet model, real-time scheduling. It ensures a real-time function based on the Real-time tasks in the model with a real-time restrict. Meanwhile, the program use of dynamic FM_PM operating systems, real-time processor to adjust to state Energy-saving way to achieve the objective during the process of implementing the mandate.

The invention discloses an extreme-low-altitude laser radar digital terrain mapping system and an extreme-low-altitude laser radar digital terrain mapping method of a small-sized unmanned helicopter. The system is formed by an airborne system and a ground station system, wherein the airborne system is used for accurately collecting data of each sensor, carrying out synchronization processing on the data of the sensors and navigation control of the unmanned helicopter, and can realize data interaction with data of the ground station system through wireless receiving and dispatching equipment; the ground station system is used for receiving and processing data from an airborne laser radar hardware system or checking and processing a data file in an offline manner; the airborne system is formed by a coprocessor and a main processor; and the coprocessor is used for collecting gesture data and position data of the system or realizing the navigation control of the airborne system. According to the extreme-low-altitude laser radar digital terrain mapping system, a mode of matching the coprocessor with the main processor is utilized so that the requirements of the instantaneity and the stability of the navigation control of the small-sized unmanned helicopter are guaranteed, the system is flexible and the function of the airborne system is conveniently expanded; and the system has the advantages of low power consumption, low cost, high integration, high efficiency and the like.

Using packet switching, the method of packet voice communication can be applicable to united public network in next generation. New layer of data link layer protocol - data link procedure (DLP) is introduced between physical transport network and voice service layer. Using current telephone number as address of voice signal packet i.e. destination address code and source address code of DLP frame realizes forwarding and exchanging voice signal in two layers. Using related signaling implements establishment and expiration of call connection as well as outputs packetized voice signal with certain length periodically. Using DLP carries out signaling information of establishment and expiration of call connection and voice signal packets output periodically. Using safety mechanism, network control management mechanism, flow management mechanism provided by DLP realizes safety management, control management and flow management of network.

The invention discloses a token bucket algorithm-based satellite data ground transmission network flow control system, which comprises a token generation module, a token placement module, a token bucket module, a data caching module, and a token request module, wherein the token generation module is used for generating tokens matched with a corresponding flow type according to a preset token arrival rate; the token bucket module comprises n token buckets for storing the tokens; the token bucket module places the generated tokens in corresponding token buckets according to a token placement strategy or discards the tokens when all the token buckets are full; the data caching module stores n flow types, and when any flow type is generated, the token request module sends data packets in the flow type according to a token request strategy or the data packets are stored in the data caching module when no tokens exist. Thus, real-time data can be preferentially sent, a bandwidth as large as possible can be ensured for non real-time data, the transmission performance is optimized, the transmission jitter is reduced, and the sending flow can be controlled as accurately as possible.

The invention relates to a video image mosaic method for adaptive light compensation, belonging to the field of video image processing. Including collecting the video shot by the camera, extracting the image pair of the first frame of the video, extracting the feature points of the video image, finding the matching feature vector pair, eliminating the mismatching feature point pair, calculating the transfer matrix, adaptive light compensation, and video stitching . The advantage is that it can realize seamless, large viewing angle, and real-time video stitching, and can be widely used in fields such as video conferencing, video surveillance, and robot vision.

The invention discloses a disposal for process and interruption in a SMP system. The method adapts separation interrupt handling and process handling on processor grade, processes specially the interruption by using one or a group processors, furthermore, monitors the process and the interruption handling, and dynamic adjusts the number of processors in two group processors according to the load index of the process and the interruption handling. The invention also discloses a handling device for process and interruption in a SMP system. The device comprises interrupted handling module, a process processing module, a process and interruption monitor module. The device ensures the process handling and interruption handling mutual noninterference to improve the hit rate of cache and to ensure the stabilizing ability and actual time character of the system.

The invention relates to a multi-sensor system for an inspection robot and an inspection method. The multi-sensor system is characterized in that a sensor system acquires the current state of the inspection robot and external environmental information and sends the acquired current state and external environmental information to a ground base station; and after receiving the current state of the inspection robot and the external environmental information, which are sent by the sensor system, the ground base station sends a control command to the inspection robot. The problem that comprehensive and accurate environmental information of a power transmission line are very difficult to acquire in the driving process of the inspection robot is solved, and autonomous line catching and dynamic centroid adjustment in the obstacle crossing process of the inspection robot are guaranteed.

An ultrasonic excitation method for eliminating the crosstalk of a robot ultrasonic ranging system adopts the pseudo-random pulse position modulation mode for constructing a sole transmitting sequence for each ultrasonic transducer and uses the genetic optimization algorithm for optimizing the time interval between the adjacent pulses in the pseudo-random pulse position modulation sequence, thus allowing the duration of the transmitting sequence of each ultrasonic transducer to be 1.5 to 2ms. The ultrasonic excitation method comprises the following steps: 1. a string of pulse signals X are generated; 2. K groups of different pseudo-random series Psi are generated; 3. the pulse positions in the pulse signals X are respectively modulated by using the K groups of pseudo-random series, thus obtaining K groups of pseudo-random pulse position modulation sequences; 4. the best pseudo-random pulse position modulation sequence of the K groups is obtained; 5. K groups of ultrasonic transducers are respectively excited. The ultrasonic excitation method can lead each ultrasonic transducer to have the sole transmitting sequence, thus fundamentally avoiding the occurrence of the crosstalk; the real-time requirement of the robot navigation can be ensured; the working efficiency of a robot ultrasonic obstacle avoidance system can be effectively improved; and the reliability of the robot ultrasonic obstacle avoidance system can be greatly improved.

The invention relates to a transmission line external damage early warning method and system based on depth learning, Firstly, a camera and a wireless communication module are mounted on more than oneUAV. Each UAV patrols the transmission line according to the preset route, height and shooting angle, and transmits the images taken in the target area to the server of the corresponding patrol station. Then the server of the patrol station builds the depth neural network for the image data transmitted by the UAV in real time by using the depth learning method, identifies the target intelligently, judges the location coordinates of the external force damage risk, and uploads it to the central console; the server of the patrol station builds the depth neural network for the image data transmitted by the UAV in real time by using the depth learning method. Finally, the central console receives the warning coordinates from each patrol station to alert the staff. The invention realizes a large range of transmission line anti-external force damage early warning by a UAV and a depth learning algorithm popularized in the market.