52results about How to "Verify robustness" patented technology

The invention relates to an MATV-feature-based face living-body detection method. The method comprises: a training set video is read and decomposed, a face region of each frame is extracted and is processed by normalization for inputting, a magnitude-angle-total-variation (MATV) optical flow filed feature based on a magnitude-angle variational method is employed, the feature is coded by using a statistical histogram to obtain an MATV histogram feature, and then the MATV histogram feature is inputted into an SVM classifier to carry out training and an SVM model is established; and for a face video frame sequence of a test set, face detection is carried out on each frame, the face region of each frame is extracted and normalization is carried out to obtain a test set face frame sample, an MATV histogram feature of the face frame sample is calculated and then is inputted into a trained SVM model, and a face living-body detection result is predicted. According to the invention, on the basis of the traditional magnitude-feature-based optical flow method, the angle feature is added and the direction characteristic of the optical flow field is considered, thereby obtaining more living-body and non-living-body face motion details. Therefore, the accuracy of the face living-body detection is improved.

The invention discloses a 422 bus solid-liquid power aircraft flying control testing system. The 422 bus solid-liquid power aircraft flying control testing system comprises a main control testing platform, a portable measurement and control box, a flying control computer, an emulation computer and an interface machine; the main control testing platform provides a human-computer interaction interface and sends a testing command to a flying control computer; the portable measurement and control box outputs the testing command to the flying control computer after performing isolation transferring on the testing command signal to achieve flying control computer power supply, analog acquisition output and switching value signal input and output channel functions; the flying control computer performs sub-system testing and control to perform flying control algorithm resolving and generate into rudder deflection control quantity; the emulation computer simulates the aircraft to perform model real-time resolving; the interface machine perform data fusion and sending. The 422 bus solid-liquid power aircraft flying control testing system can achieve full-system semi-real-object testing, comprehensive monitoring of bus information, real-time storage and display of data, obtains possible conditions during flying to examine in advance and lays foundation for smooth execution of flight tests.

The invention relates to an object recognition method based on surf. According to the method, a feature vector set is set up by using SURF of multiple images of an object undergoing illumination, size, rotation and noise changes, the feature vector set is matched with a feature vector set of a standard image, and then a final recognition result is determined by adopting a weight similarity product method. The object recognition method is high in recognition rate, and robustness of the object recognition method is verified.

The invention discloses a designing method of a high-efficiency high-coverage-rate function test case, which sequentially comprises the following steps: analyzing a test object and a test requirement; starting the case; analyzing an expected result of the case; designing a forward function teat case according to the teat object and the requirement analysis when a program receives and processes an input variable; realizing functions for testing software requirements; then, completing the case by using various case designing methods; designing a reverse abnormal condition case according to the teat object and the requirement analysis when the program does not receive the input variable; and completing the case by using various case designing methods and estimating a test result in advance. The designing method of a high-efficiency high-coverage-rate function test case has the advantages that the teat case is designed through positive thinking and reverse thinking, the positive thinking based on a software function requirement is realized so as to ensure the correct realization of the software function, the reverse thinking based on the abnormal input, the abnormal operation, and the like is realized so as to test the robustness of the software and ensure the high coverage rate of the test, and the execution effect is quicker and more accurate.

The verification system for fast autonomous deep-space optical navigation control prototype belongs to the field of semi-physical simulation system based on virtual reality. It includes the autonomous optical navigation control and dynamic simulation system to be verified with the selected landed planet and area to generate the navigation information and control command of deep-space detector to complete the dynamic simulation of the deep-space detector; a target simulating system to generate the 3D virtual environment the deep-space detector faces and to be shown in the display based on the detector information produced through dynamic simulation; and an optical navigation camera system to take the image on the display and to drive the generation of next step of deep-space detector navigating information and control command and complete the dynamic simulation of the deep-space detector.

The invention belongs to the field of aerospace craft guidance, and aims to enable an RLV to still obtain a guidance instruction enabling a track to strictly meet heat flow, overload and dynamic pressure constraints under the condition that the RLV is affected by external disturbance and uncertain factors, and finally land at an appointed runway position at a proper speed. Therefore, the technicalscheme adopted by the invention is as follows: the full-stage reentry return guidance method for a reusable launch vehicle comprises the following steps of: 1, guidance of an initial descending section; 2, gliding section guidance; 3, guidance of a terminal energy management section; 4, automatic landing section guidance: keeping the heeling angle to be 0 in the transverse direction; in the longitudinal direction, the automatic landing section is designed to be formed by combining a quasi-balanced gliding section and an exponential gliding section, an attack angle instruction is determined bydesigning a track angle function, and the height of a switching point of the two sections is designed through a secant method, so that the RLV can make contact with the ground at a preset speed and atrack angle. The method is mainly applied to aerospace craft guidance occasions.

The invention provides a simulation logging data generation method and system, a computer device and a computer readable storage medium, and relates to the technical field of intelligent well drilling. The system comprises a logging data acquisition module used for acquiring logging data; a matrix data determining module which is used for obtaining matrix data through a wide-angle eye mechanism according to the logging data; and an analog data generation module which is used for generating analog logging data according to the matrix data based on the generative adversarial network. The methodis based on a small amount of real logging while drilling data, takes a certain randomness into consideration, and a large number of different logging data can be rapidly generated, so that an effective data set is provided for a large number of pilot drilling algorithm models, the robustness of the algorithm can be verified, the blindness of the model in practical application is reduced, the riskis reduced, a guiding effect is achieved on production and environment analysis, and the great economic and social benefits are achieved.

The invention provides a network protocol safety test method and device, electronic equipment and a storage medium. The method comprises the following steps: establishing a plurality of protocol models of a plurality of protocol layers of a protocol set based on an original message of the network protocol set; generating a plurality of variation messages according to the plurality of protocol models and variation strategies; sending the variation message and the background flow message to the equipment to be tested; and monitoring a transmission process of the to-be-tested device in receivingthe variation message and the background flow message so as to judge whether the to-be-tested device is safe or not.

The invention discloses a pi-network-based millimeter wave frequency band receiver with an electrostatic discharge protection function, and belongs to the field of radio frequency and millimeter wave integrated circuit designing. The receiver comprises an electrostatic discharge (ESD) protection circuit consisting of back biased diodes, a low-noise amplifier (LNA) in a two-stage amplification structure, mixers of branch circuits I and Q, gain variable amplifiers of the branch circuits I and Q, a balun and an output ESD protection circuit consisting of the back biased diodes. Connection relationships among the circuits are that: the ESD protection circuit is connected in series with the LNA; the other end of the ESD protection circuit is an input end; the other end of the LNA is connected with the common input end of the two mixers of the branch circuits I and Q; the input end of the balun is a local oscillation (LO) end, and the output end of the balun is connected with the two local oscillation signal input ends of the mixer of the branch circuit I; the output ends of the mixers of the branch circuits I and Q are connected with the input ends of the gain variable amplifiers of the two branch circuits respectively; and the output ends of the gain variable amplifiers are connected to the final output end of the circuit after passing through the ESD protection circuit. By the invention, the electrostatic reliability of the circuit of the receiver can be ensured, and the overall performance of broadband input matching, high gain and low noise of the receiver can be improved.

The invention provides a sleep staging method based on BLSTM and a sleep staging device on the BLSTM. The BLSTM comprises two training layers and one output layer, wherein each training layer comprises 16 units, the output layer comprises 4 units, and the four units correspond to an awakening stage, a light sleep stage, a deep sleep stage and a rapid eye movement stage of sleep respectively; and input characteristics of the BLSTM comprise a characteristic extracted from an RR interphase signal, a characteristic extracted from a breathing signal and a characteristic extracted from cardiopulmonary coupling. The method and device acquire cardiopulmonary signals of a testee using a low-cost wearable multi-sensor system, conduct four-classification sleep staging by a bidirectional recursive neural network model which has an LSTM function and realize accurate staging of sleep.

The invention discloses a target tracking method and device. The method comprises the following steps: extracting image features from an image target candidate region of a video frame, and outputtingan initial response value; smoothing an initial response value of the video frame; and updating the response value, and determining the position information of the target in the video frame. The technical problem that the target tracking method is poor in effect is solved. For the output response of a network, scale transformation is often caused; according to the method, the phenomena of multi-peak, drifting occur due to complex target states such as sudden motion and the like, the similarity relation between adjacent super-pixel blocks of the image is fully considered, the initial response value is constrained by utilizing the relation, the drifting phenomenon in the tracking process is inhibited, and then the position information of the target is corrected. And therefore, the tracking precision and the tracking efficiency of the target tracking method are improved.

The invention discloses a method for deducing an IP network congestion link packet loss rate range. The method comprises the following steps: creating a to-be-tested IP network Bayesian network model, learning each link congestion prior probability to serve as positioning congestion link experience and knowledge, and positioning a congestion link by Bayesian maximum posterior; and providing a path collection method relative to clustering congestion link and having similar performance, solving coefficients with similar performance in a clustering path set, and deducing the congestion link in the entire to-be-tested IP network and the link packet loss rate range thereof by means of the greedy heuristic search cycle. According to the method, the link which is subjected to the congestion most easily in the congestion paths by the link congestion maximum posterior cost values Cp through the link congestion BMAP criterion, the traditional experience method using a bottleneck link as congestion link positioning is replaced, and thus the congestion link positioning accuracy is improved; and the congestion link reduction error caused by the minimum link coverage theory is effectively avoided by the clustering path set, and the deduction performance of the algorithm is improved.

The invention discloses a test system and method for protecting an open port from attacking for network equipment, relating to the technical field of computer network security. The method comprises the following steps of performing ping connection on an internet protocol (IP) address of a local area network (LAN) side of to-be-tested equipment by a client side, and performing ping connection on an IP address of a wide area network (WAN) side of the to-be-tested equipment by a server; running port scanning software on the client side, scanning the IP address of the LAN side of the to-be-tested equipment, recording a scanned first open port number, then running attack software on the client side, performing different types of message attack on the first open port number, and recording a test result; and running the port scanning software on the server, scanning the IP address of the WAN side of the to-be-tested equipment, recording a scanned second open port number, running the attack software on the server, performing different types of message attack on the second open port number, and recording a test result. The method is better in pertinence.

The invention discloses a Harris eagle optimization algorithm based on random traceless Sigma point variation. The Harris eagle optimization algorithm comprises the following steps: generating initialpositions of N eagles in a feasible region by utilizing a random initialization method; recording the position of the optimal eagle, and carrying out random traceless Sigma point variation; updatingescape energy by using a nonlinear strategy; selecting and executing corresponding exploration and development strategies according to the escape energy and the random numbers, and using quasi-reflection and quasi-reverse solution updating in the development strategies according to conditions; calculating fitness values of eagle population individuals, and determining a current optimal solution and an optimal target value according to an optimal principle; repeating the steps S2-S7 until the maximum number T of iterations is reached; and returning the optimal solution and the optimal target value. The quasi-reverse and quasi-reflection learning strategies enhance the diversity of the population and accelerate the convergence rate of the algorithm; logarithm nonlinear convergence factors are used for balancing algorithm exploration and development capability; and a Sigma point strategy is randomly symmetrical, so that the algorithm is prevented from falling into local optimum.

The invention discloses a test method and system for USB (Universal Serial Bus) storage stability of a route. The method comprises the following steps of receiving a test file, wherein the test file is stored in an external storage device of a tested route; receiving a USB interface, connected to the tested route, of external storage equipment; and repeatedly reading and writing the test file in the external storage equipment. By the method, the purposes of verifying the stability and the robustness of a USB storage system of the route can be achieved.

The verification system for fast autonomous deep-space optical navigation control prototype belongs to the field of semi-physical simulation system based on virtual reality. It includes the autonomous optical navigation control and dynamic simulation system to be verified with the selected landed planet and area to generate the navigation information and control command of deep-space detector to complete the dynamic simulation of the deep-space detector; a target simulating system to generate the 3D virtual environment the deep-space detector faces and to be shown in the display based on the detector information produced through dynamic simulation; and an optical navigation camera system to take the image on the display and to drive the generation of next step of deep-space detector navigating information and control command and complete the dynamic simulation of the deep-space detector.

The invention provides a lithium battery residual life prediction method based on an MFF multi-core GPR algorithm, and the method comprises the following steps: obtaining a battery data set, taking the battery capacity as a lithium battery life prediction index, carrying out the extraction of a plurality of features of a charging and discharging curve of the battery data set, and forming a plurality of single feature sets; performing weighted linear combination on the extracted features, and calculating the combined features; preliminarily selecting a plurality of single kernel functions according to trend characteristics of the feature indexes, and performing weighted linear combination on the kernel functions to obtain a combined kernel function; dividing the data set into training data and test data, taking the combined features obtained in the step 2 as input features of the GPR algorithm, substituting the combined kernel function obtained in the step 3 into the GPR algorithm, training a GPR model on the training data set, then performing prediction, and stopping prediction until the predicted battery capacity reaches a failure threshold value, and calculating the residual life value RUL of the lithium battery according to the time when the failure threshold value is reached.

The invention discloses a monocular vision-based lane line detection and tracking method. The method comprises the following steps of: A, image acquisition; B, image preprocessing; C, detecting whether the image is a first frame image, if yes, sending the image into a first frame image lane line detection module, and if not, sending the image into a tracking process lane line detection module; andD, detecting output results of the first frame image lane line detection module and the second frame image lane line detection module, determining whether the first frame image lane line detection module and the second frame image lane line detection module are the last frames, if yes, ending, and if not, returning to the tracking process lane line detection module for reprocessing. According tothe invention, the research work of lane line detection can be completed. And the lane line detection tracking work of the video stream is completed, and the real-time performance and robustness of the video stream are verified.

The invention discloses a visual positioning hardware-in-the-loop simulation platform and simulation method. The visual positioning hardware-in-the-loop simulation platform comprises an automatic driving simulation system, a darkroom scene system and a visual positioning system, the visual positioning system comprises a visual positioning calculation platform and a camera, and the visual positioning calculation platform is in communication connection with the automatic driving simulation system; the visual positioning calculation platform obtains virtual sensor data and a virtual scene high-precision map from the automatic driving simulation system; and the darkroom scene system comprises a darkroom, a high-definition display screen and a camera in the visual positioning system are placed in the darkroom, the high-definition display screen is connected with the automatic driving simulation system through a video line, the camera is placed in front of the high-definition display screen, and the camera collects videos displayed on the high-definition display screen by the automatic driving simulation system in real time. The visual positioning computing platform and the camera are connected into the system, hardware-in-the-loop simulation is achieved, and the real drive test condition is simulated to the maximum extent.

The embodiment of the invention discloses a fixed time terminal sliding mode robust fault-tolerant control method and device for a liquid-filled spacecraft. The method comprises the following steps: constructing a coupling kinetic model of a liquid-filled spacecraft based on a viscous spherical pendulum model of liquid sloshing; and designing a robust adaptive fault-tolerant control strategy for the coupling dynamics model based on a fixed time control strategy and an adaptive control algorithm. According to the scheme provided by the invention, the singularity problem existing in a terminal sliding mode control scheme can be overcome, and meanwhile, the state of the liquid-filled spacecraft is ensured to be quickly converged to a fixed-time sliding mode surface.

The invention discloses a high-throughput sequencing method for efficient research of RNA interaction interactome and application of the high-throughput sequencing method. The method includes steps: (1) arranging DNA probes at intervals of 100bp according to length of to-be-tested RNA; (2) after crosslinking fixation with 1-3% (v / v) of formaldehyde solution, subjecting cells to ultrasonication toobtain cell lysis solution in nucleic acid length range of 100-500nt; (3) adding the DNA probes for hybridization, gathering target RNA, washing, and purifying to obtain RNA fragments; (4) connecting3' terminals of the RNA fragments with RNA joints, performing product reverse transcription to obtain cDNA, connecting the 3' terminal of the cDNA with a DNA joint to obtain a library of the target RNA and RNAs in interaction with the target RAN; (5) amplifying the library, and carrying out detection analysis. The method can be used for library construction of low-content interaction RNAs.

The invention relates to a load frequency control device suitable for an information uncertainty system, which comprises an information acquisition module, an optimization decision module and an adaptive control module. The information acquisition module is used for accessing a state signal of a communication system at the moment N into the device through sensor sampling measurement, and integrating and uploading a sampling signal to an agent node; the optimization decision module receives the information output by the information acquisition module, performs synchronous sampling conversion according to the occurrence probability measure of the random event of the communication system, analyzes the information uncertainty degree of the system through the established model of the information entropy H, and inputs the analysis result into the adaptive control module; the adaptive control module inputs a data analysis result into the adaptive control module designed based on a fuzzy rule,aiming at the probability measure representation of each quantity influencing the uncertainty of the system information, a mathematical model of information entropy is established by integrating eachquantity of time lag, packet loss and error codes of an information system, the modeling of the uncertainty of the system information is completed. A system load frequency model is established by quantizing information uncertainty, and the frequency control considering information uncertainty is realized.

The invention discloses a low-dose double-tracer PET (positron emission tomography) reconstruction method based on an attention mechanism, namely, a low-dose double-tracer PET image reconstruction problem is solved by utilizing a convolutional network model based on the attention mechanism, and standard dose estimation and double-tracer PET signal separation are realized in a sinogram. According to the method, by means of a powerful feature extraction tool of deep learning, a single tracer PET image of a standard dose can be reconstructed from a low-dose double-tracer PET sinogram, and the method has robustness for individual differences, tracer combinations and sampling protocols.

The invention discloses a cladding pool morphology recognition and closed-loop control method based on instance segmentation. The cladding pool morphology recognition and closed-loop control method comprises the following steps of initializing arc welding additive manufacturing output power and output power variable quantity; collecting a color image of the arc welding additive manufacturing cladding pool; updating arc welding additive manufacturing output power; segmenting a cladding pool in the cladding pool color image, and meanwhile, generating an instance mask of the cladding pool and pixel coordinates of the candidate frame to complete instance segmentation of the cladding pool; taking the pixel coordinates of the candidate boxes for representing the number of pixels of a long axis and the number of pixels of the fusion width, calculating the variation of the number of pixels of the fusion width, and feeding back the generated output power variation; and judging whether the current cladding pool color image is the last frame or not. The method is good in real-time performance and high in recognition precision, when molten drops cover the edge of the cladding pool, accurate detection and recognition can be achieved, the effectiveness and robustness of the algorithm are verified, the morphology of the cladding pool is detected with low cost and high efficiency, and a reliable basis is provided for real-time evaluation of arc welding additive manufacturing.

The invention discloses a model body and a calibration method suitable for geometric correction of a general trajectory cone beam CT system. The calibration method comprises the following steps: constructing back projection line coastal introduction auxiliary parameter virtual intersection points of a small ball at different angles based on projection data of the small ball; constructing a target function of the back projection line intersection degree according to the virtual intersection point; expanding the objective function to obtain a sum objective function of the N small balls; the optimal solution of the sum objective function is solved through an optimization algorithm, and initial three-dimensional coordinates of the mass centers of the N small balls are obtained; and registering and averaging the initial three-dimensional coordinate values of the N small balls based on a rigid motion registration algorithm of singular value decomposition to obtain an accurate three-dimensional coordinate value which is used as a reference value of the geometric calibration method of any track. The method aims at accurately obtaining high-precision space parameters of the die body without depending on machining precision, and further provides a calibration die body for geometric correction of a common track cone beam CT system, especially a high-resolution cone beam CT system.

The invention provides a planetary detection capture brake and device separation all-physical simulation experiment device and method, and belongs to the field of aircraft ground simulation. A groundmonitoring and control system is connected with other systems through cables, a position measuring system returns data of an encoder to the ground monitoring and control system, and the ground monitoring and control system sends a control command of a servo motor to a motion simulation system; a camera controller in a visual measurement system is in communication with the ground monitoring and control system through an RS422 serial port and transmits position information feedback of a motion simulator; the visual measurement system is connected with a simulated gravity generation system through a cable, the visual measurement system collects images of the motion simulator on the simulated gravity generation system, and the motion simulation system is fixedly connected with the simulated gravity generation system. The device and method have the advantages that superiority which mathematical simulation cannot be incomparable to, and unpredictable problems can be met, so that the reliability and robustness of a control algorithm are verified, and the reliability of engineering implementation is improved.

The present disclosure provides functional test equipment for a device under test and method of testing the device under test. The functional test equipment includes a first power supply, a second power supply and a relay system. The first power supply is configured to generate a first supply voltage. The second power supply is configured to generate a second supply voltage different from the first supply voltage. The relay system is configured to electrically couple the first power supply or the second power supply to the device under test, wherein the first supply voltage is applied to the device under test for a first duration and the second supply voltage is applied to the device under test for a second duration less than the first duration.

The invention discloses an ignition algorithm verification test device for an air bag restraint system controller, which can replace a trolley test for verification of impact strength and ignition algorithm of the air bag restraint system controller in early stage of development of the air bag restraint system controller. The test device comprises four modules, including a test stand rack, a pendulum bob impact acceleration mechanism, a test pulley component and a data acquiring module; the test stand rack comprises a main test stand rack, a pulley slide way and a pulley buffering device; the test stand rack acts as a component for bearing load in the test; the pendulum bob impact acceleration mechanism comprises a handheld electromagnet, a pendulum bob, a pendulum rod vernier device, a pendulum rod and an oscillation angle measurer; the pendulum bob impact acceleration mechanism acts as a power output component in the test; the test pulley component comprises a pulley, an energy absorption cylinder fixing seat and an energy absorption cylinder; the test pulley component acts as an automobile for simulating impact and a carrier of an air bag restraint system controller sample; and the data acquiring module comprises an acceleration sensor and a speed measurer and is used for acquiring the test data in the test. The algorithm verification test device disclosed by the invention has the advantages of simple structure, low cost, convenience in operation and high efficiency, etc.

The invention discloses a horizontal set image segmentation method fusing uneven local gray scale and local variance. A problem that segmentation effects of images with uneven gray scales should be improved in the prior art is solved. The method comprises following steps of 1, establishing a mathematic model with the uneven gray scale for features of an image with the uneven gray scale; 2, drawing image gray scale distribution through a Gauss distribution function; 3, processing an original image through low-pass filtering; 4, establishing an image local data item according to the step 3, establishing a regularization item simultaneously and constructing an overall level set energy function by combining a local statistical item established in the step 2; and 5, converting the energy function into a partial differential equation according to the Euler-Lagrange theorem, thereby achieving minimization of the energy function through gradient descent. According to the invention, in the improved model with the uneven gray scale, difference between the original image and an estimated image is considered, problems in the improved model with the uneven gray scale of the traditional image is analyzed and segmentation results are improved after the difference is introduced.

The invention discloses a train ground automatic neutral section passing method and system, terminal equipment and a storage medium. The method comprises the steps: obtaining a train operation position, and determining a power supply for supplying power to a train according to the operation position; specifically, when the train is located in a transition region, according to the voltage and current of the transition region and the parameters of a quadruple auxiliary converter device, determining an output voltage decoupling modulation signal of the quadruple auxiliary converter device in the transition region and an output voltage modulation signal of the power supply arm corresponding to the transition region; and when the train is located in the electrical isolation region, pre-modulating the voltage amplitude and phase of the quadruple auxiliary converter device by taking the voltage of the power supply arm B as a target. According to the invention, over-voltage and over-current impact influences of pantograph-catenary arcs on a vehicle-catenary system can be avoided; flexible voltage switching between adjacent power supply arms can be realized, and short circuit and transient impact influences caused by direct voltage switching can be effectively avoided and suppressed; and in addition, the rated operation condition of the train is guaranteed through alternate transfer of the power, and the speed loss of the train is reduced.