56results about How to "Meet the needs of real-time detection" patented technology

The invention relates to an image detection method for detecting foreign matter at the bottom of a vehicle. The method sequentially includes the following steps of image input, image registration, binary image acquisition and foreign matter detection. The method can be used for fast recognizing whether the bottom of the vehicle going in or out of a barrier gate carries foreign matter or not, a vehicle chassis image provided by a vehicle chassis imaging system is subjected to signal analysis and processing, a real-time reliable foreign matter detection method is provided for detection of the chassis of the vehicle, and therefore the foreign matter at the bottom of the chassis of the vehicle can be fast, accurately and efficiently detected, the detecting accuracy is improved, interference of manual operation and environment changes is eradicated, and the requirement for real-time detection of the bottom of the vehicle in the security and prevention market in China is met. The time needed for recognition is no longer than five seconds, and the recognition efficiency is high.

The invention relates to an FPGA (field programmable gate array)-based power harmonic analyzer, belonging to the technical field of power electronics. The power harmonic analyzer comprises a sensor circuit and an AD sampling circuit as well as a data acquisition module, a data caching module, a data type conversion module, a windowing module, an FFT (fast Fourier transformation) computation module, an operational result storing and interruption generating module, a Nios II soft core processor module and an LTM (long-term memory) display module which are arranged in an FPGA. In the invention, harmonic acquisition and processing at the front end are realized by adopting hardware, thus the acceleration advantage of hardware is fully played; and the processed data is further calculated and controlled by utilizing the Nios II soft core processor in the FPGA, the process is completed on an on chip bus of the FPGA, data transmission is realized, and speediness of hardware design and control flexibility of the Nios II soft core processor are given full play.

The invention provides a flame detection method based on image target detection, which belongs to the field of image processing, fire detection and video monitoring. The method comprises steps: firstly, a flame detection data set containing flame images and annotation information of each image is built, and the flame detection data set is divided to a training set and a test set; a deep convolution neural network model is built, the training set is used to carry out iterative updating on the model, the test set is used to calculate a loss function for the updated model, and if the loss function for the current model does not drop any more, the model completes training; and a real-time video is photographed, the model which completes training is used to detect each frame of image, if flameexists, the coordinate position of the flame in the image is outputted by the model, and a rectangular frame is used for marking. The phenomenon that features are manually designed for generating a candidate area for suspected flame is not needed, the deep convolution neural network model can be directly used for carrying out flame detection on the whole image, the position information of the flame is obtained, early warning of a fire is thus carried out, and hazards brought by fire can be reduced maximally.

The invention provides a word sequence-based phishing URL detection method and system, and aims at detecting Phishing URL. The method comprises the following steps of: carrying out word segmentation on a URL character string so as to obtain vector expression of a word sequence; automatically learning context information and features in the word sequence by utilizing a deep learning model without artificially extracting word related text features included in URLs; and detecting Phishing URLs by adoption of the trained model. According to the method and system, problems in word feature-based phishing URL detection are solved.

The invention discloses a method for detecting a road traffic incident based on a double-section annular coil detector. The conventional method is focused on traffic parameter analysis and is used for judging a traffic incident by setting corresponding thresholds. The method for detecting the road traffic incident based on the double-section annular coil detector comprises the follows: a step of obtaining traffic flow data in each sampling interval; a step of calculating relative change rate and evaluation standard deviation of traffic flow parameters between two adjacent detection section based on the pre-treated traffic flow parameters; a step of calculating a judging index of a double-section road traffic incident based on the calculated relative change rate and the evaluation standard deviation; and a step of determining whether the traffic incident occurs based on the judging index of the double-section road traffic incident. The method for detecting the road traffic incident based on the double-section annular coil detector, disclosed by the invention, can lighten the influence of fluctuation of traffic flow parameter of single detection section to detection result and improve detection accuracy.

The invention relates to a fatigue driving early warning system and method based on a depth camera. The fatigue driving early warning system comprises a depth camera, a feature point positioning module, a state recognition module and a driving state early warning module. The method comprises the following steps: acquiring an infrared image sample and a depth image sample, and acquiring a human face region and a human face feature point set; based on the human face feature point set acquired in the step S2, performing calculation to obtain a corresponding classification state and fatigue stateevaluation. Aiming at the night recognition feature extraction problem, the system and the method adopt a depth camera to acquire an infrared image and a depth image, the eye state is judged accordingto the eye length-width ratio of front face feature points or the upper and lower eyelid distance of side face feature points, and the fatigue state is comprehensively evaluated according to the eyestate and the mouth state. The method can meet the requirements of real-time detection, and has a practical value for fatigue detection and early warning.

The invention provides a multispectral image acquiring method on the basis of line scanning hyperspectral imaging. The method comprises the following steps of: presetting more than one characteristic wavelength for a sample to be detected; determining a relation between line numbers on a spectral axis of a camera and the wavelengths; by utilizing the more than one characteristic wavelength of the sample to be detected and the relation between the line numbers on the spectral axis of the camera and the wavelengths, calculating more than one line number on the spectral axis of the camera, which corresponds to the more than one characteristic wavelength; carrying out acquisition on the sample to be detected on a conveying device by a hyperspectral spectrometer and the camera, extracting line pixel information corresponding to the more than one line number from a two-dimensional matrix obtained by acquisition each time, and combining the line pixel information corresponding to the more than one line number in a plurality of two-dimensional matrices obtained by acquisition so as to obtain a multispectral image under multiple characteristic wavelengths. The method can implement acquisition of the multispectral image under the specific wavelengths.

The invention relates to the technical field of risk control of the Internet financial industry, particularly to a fraud behavior detection method based on whale algorithm optimization LVQ neural network. The method comprises six steps. According to the invention, compared with BP, RBF, SOM and other neural networks, the LVQ neural network has the advantages of simple structure, short training time, stronger non-linear classification processing capability and the like; compared with optimization algorithms such as a genetic algorithm, a particle swarm algorithm and the like, the whale optimization algorithm has the characteristics of simple parameter setting, strong function optimization capability, strong global optimization capability, good convergence stability and the like; and the initial weight of the LVQ neural network is optimized by using the whale optimization algorithm to improve the global fitting capability, the learning rate and the prediction precision of the LVQ neuralnetwork, so that the requirement of real-time detection of networked financial fraud behaviors can be met.

The invention discloses a double-band imaging discharge detector for a high-speed train, and an application method of the double-band imaging discharge detector. The detector comprises a visible light lens, a visible light camera, an ultraviolet lens, an ultraviolet camera, and a processing and control board, and a power board. The detector is installed behind a pantograph of the roof of a locomotive. The visible light camera matched with the visible light lens obtains visible light background images of the pantograph and an overhead line system in front of the locomotive, and transmits the images to the processing and control board. The ultraviolet camera matched with the ultraviolet lens and a filter obtains the ultraviolet target images of the pantograph and the overhead line system in front of the locomotive, and transmits the images to the processing and control board. The processing and control board carries out the image processing, and judges whether the key parts of the pantograph and the overhead line system perform partial discharge and whether the pantograph and the overhead line system generate arcs or not, and carries out the positioning and displaying of video images when the partial discharge happens or the arcs are generated. The detector can timely find out the partial discharge conditions, detects the arc state in real time, and reduces the unnecessary economic loss in order to guarantee the normal transportation order of the high-speed train.

The invention relates to a planar object detection method based on a continuous edge characteristic, and belongs to the technical field of machine vision image processing. The method comprises the steps including extracting the continuous edge characteristic, establishing a target contour model, extracting candidate objects based on a Hough transformation idea, screening the candidate objects and realizing accurate positioning. The method is suitable for an industrial scene, and can be used to establish a model of an unknown planar object rapidly and realize object detection. The method needs not to be adjusted for the specific object, and is suitable for different planar objects (as well as objects whose surfaces are planar), and the problem that the university of a present visual system is low is solved to certain extent. The detection precision is higher, the processing speed is high, and certain anti-noise capability is provided.

The invention discloses a method and device for image authentication. The method comprises the steps of 1, a matched image and an image secret key are obtained; S2, the matched image and the image secret key are converted into a corresponding gray scale single channel graph; S3, through an ORB characteristic extracting algorithm, extraction of image characteristics and matching are conducted on the converted matched image and image secret key; S4, whether an ORB characteristic detection is passed or not is judged, if yes, the step S5 is entered, and if not, matching fails; S5, template matching is conducted on the matched image and the image secret key, if matching succeeds, encoding succeeds, and if matching fails, encoding fails. According to the method and device for the image authentication, by combining the ORB characteristic detection and template matching, blind spots in image matching can be optimized, and the comprehensive image matching level can be improved.

The present invention discloses a target behavior detecting method and a server, wherein the method includes: acquiring a scene image and identifying a first target and a second target; acquiring thefirst location data of the first target and the second location data of the second target; calculating an association degree value between the first target and the second target according to the firstlocation data and the second location data; determining an associated target pair according to the association degree value, wherein the associated target pair includes the first target and the second target; extracting a target image including the associated target pair from the scene image. The target behavior detecting method and the server provided by the embodiment of the present invention simplify the target behavior detection into target recognition and target association, reduce algorithm complexity, thereby improving the calculation speed. Thus, the target behavior detecting method provided by the embodiment of the present invention can meet the needs for real-time detection.

The invention relates to a tire X-ray image impurity defect detection method. The method comprises the following steps: S1, establishing an image data set containing impurity defects; s2, enhancing the data set by using rotation, mirroring and brightness transformation; s3,building an improved YOLOv4-tiny network; s4, setting hyper-parameters and network parameters for training; s5, inputting a to-be-detected image to carry out defect detection. According to the invention, end-to-end tire impurity defect positioning is realized, the requirement of real-time tire defect detection can be met, and the production efficiency is greatly improved.

The invention relates to a chemical system multi-working-condition fault detection method based on local adaptive standardization and belongs to the technical field of chemical process monitoring, industrial data processing and process system engineering. According to the method, a local adaptive standardization method is provided, the variational automatic encoder technology of the deep neural network is applied, an average value of the data in the local moving window is calculated to serve as an average value parameter of local adaptive standardization, different average values are used fordifferent data, and the adaptive capacity is achieved. The method is advantaged in that local adaptive standardization processing is utilized, and fault detection is carried out by detecting whether the data in a local moving window deviates or not, and the method can be suitable for any working condition, has higher accuracy and stronger generalization ability, can meet the requirement of real-time detection, and avoids chemical accidents or reduces the harm caused by accidents through early warning of faults.

The invention discloses a cloth defect intelligent recognition method based on deep learning, and the method comprises the following steps: S1, carrying out the photographing of cloth with different forms of defects through a camera, obtaining a picture material, and dividing the picture material into a training set and a verification set according to a certain proportion; S2, labeling the obtained picture material; S3, inputting the labeled data into an object detection model for training, and storing model parameters obtained by training; and S4, endowing the obtained model parameters to anoriginal model, taking a test image as the input of the model, processing the input image test, carrying out feature processing on the test image by the SSD model, and then predicting the type and theposition of the defect. According to the method, the end-to-end processing effect is achieved, the working efficiency is improved, and meanwhile, the robustness of the model is effectively improved through complex feature extraction calculation in the model.

The invention discloses a real-time epileptic seizure detecting and monitoring system for epileptic video electroencephalogram examination, which comprises a data processing module, a model selection module, an epileptic seizure alarm module and an interactive collaborative labeling module, wherein the data processing module is used for processing input electroencephalogram signal data into data forms required by epileptic seizure detection models; the model selection module is used for constructing and selecting the epileptic seizure detection model for a current patient; the epileptic seizure alarm module is used for detecting epileptic seizure and giving alarms; and the interactive collaborative labeling module is used for interaction among the epileptic seizure alarm module, accompanying persons and a monitoring room. Aiming at practical problems in clinical video electroencephalogram examination, a functional module is designed, the feature learning ability and training speed of the epileptic seizure detection models and the accuracy of real-time epileptic seizure detection are improved, and the experience of accompanying persons and doctors is greatly improved.

The invention discloses a tower inclination detection device and method. Limits or restrictions of an external environment can be broken to accurately detect the inclination of a tower. The tower inclination detection device comprises an annular communicating vessel, two or more detection media and a medium interface detection unit, wherein the annular communicating vessel comprises an annular pipeline and three or more columnar tubes arranged on the annular pipeline and communicated with the annular pipeline and the outside; the detection media are accommodated in the annular communicating vessel, and interfaces between the upper-layer detection media and the lower-layer detection media are positioned in the columnar tubes; the lower-layer detection media have densities higher than those of the upper-layer detection media, and are insoluble with the upper-layer detection media; the medium interface detection unit is used for detecting the heights of the interfaces between the upper-layer detection media and the lower-layer detection media in the three or more columnar tubes.

The invention provides a real-time tracking detection method for the state of a split knife switch. The real-time tracking detection method comprises the steps: collecting a real-time monitoring videoof the split knife switch, obtaining each frame of image of the video, building a boundary line model of a knife switch arm of the type, and storing the boundary line model as a model file; acquiringa first frame image of the video, positioning a knife switch in the first frame image, and extracting tracking feature points; performing detection according to a positioning result, and calculatingan initial included angle between the left knife switch arm and the right knife switch arm; and tracking and positioning the knife switch arm in the next frame of image, estimating a system error, determining a correction included angle between the left knife switch arm and the right knife switch arm in the next frame of image, judging the real-time state of the knife switch in the next frame of image until the knife switch is detected to stop moving, and verifying the final state of knife switch movement in the video. According to the real-time tracking detection method, the problem of low real-time detection precision of the state of the split knife switch is solved.

The invention provides a malicious webpage identification and detection method based on a static field, a computer and a storage medium, and belongs to the technical field of webpage identification and detection. Comprising the following steps: step 1, monitoring webpage traffic in real time, and extracting a URL address of an HTTP head; step 2, matching the URL address with a URL address stored in a blacklist library; 3, analyzing the webpage traffic which fails to be matched; step 4, crawling JS (JavaScript) and CSS (Cascading Style Sheet) files in the analyzed webpage traffic; 5, extracting a webpage fingerprint of the target webpage; step 6, identifying webpage traffic; 7, comparing the URL addresses of the two webpages; if the URL addresses are the same, the webpage in the flow is a normal webpage, and a matching log is stored; and if the URL addresses are different, the webpage in the flow is a malicious webpage, and blocking is carried out. The technical problem that the requirement of real-time detection in practical application cannot be met is solved. The technical effect of reducing the time cost in the webpage matching process is achieved.

The invention discloses an embedded safety helmet detection method and equipment. The method comprises the following steps: collecting to-be-detected image data; performing feature extraction on the to-be-detected image data by using a feature extraction network to obtain a plurality of feature maps of different dimensions; the feature extraction network is a simplified darknet-53 neural network of which the network layer number and the convolution kernel number are less than those of a traditional darknet-53 neural network; a constructed improved YOLOv3 network model is used to detect and identify the plurality of feature maps of different dimensions, and a safety helmet image region is obtained; wherein the improved YOLOv3 network model comprises four layers of YOLO detection layers. According to the invention, the network parameter quantity can be reduced, the detection speed is improved, and the precision is higher than that of the original YOLOv3.

The invention provides an efficient multi-scale sheep front face detection method. The method comprises the following steps: S1, acquiring data set images containing different scales of sheep front faces; s2, labeling a sheep face frame and seven face feature points in the data set image of the sheep front face, and obtaining a labeled data set; s3, randomly dividing the labeled data set into a training set, a verification set and a test set; s4, inputting the training set into an efficient face and feature point detection network for training to obtain a trained network model; s5, preprocessing a to-be-detected picture and a video key frame, and inputting the preprocessed to-be-detected picture and video key frame into the trained network model to obtain a sheep face detection result; and S6, screening the detected sheep front face according to whether seven face feature points of the sheep front face are detected or not and whether the detection frame and the seven face feature points meet the positive angle feature relationship or not.