74 results about "Discrimination threshold" patented technology

The invention relates to an abnormal image detection method based on a supervised generative adversarial network. The method comprises the following steps: 1) obtaining a to-be-detected picture; 2) inputting the to-be-detected picture into a trained Supervised GANomal model to obtain a corresponding abnormal category trust value; 3) judging whether the abnormal category trust value is greater thanan abnormal category discrimination threshold, if yes, judging the abnormal image as an abnormal image, and if not, judging the abnormal image as a normal image; Wherein the Supervised GANomain modelcomprises a generation network, a coding network, a discrimination network and a classification network, the generation network, the coding network and the discrimination network are used for learning feature distribution of a normal sample, and the classification network is used for distinguishing the normal sample from an abnormal sample. Compared with the prior art, the method has the advantages that the normal sample and the abnormal sample can be effectively distinguished, the robustness is good and the like.

An apparatus for discriminating between cardiac events that includes an input circuit receiving signals indicative of the cardiac events, a first output circuit generating pulses in response to the received signals, and a microprocessor determining whether a signal received by the input circuit subsequent to the generated pacing pulses corresponds to a predetermined cardiac event in response to an elapsed time period between the generated pulses and the subsequently received signal and a predetermined discrimination threshold.

A cryptographic key distribution method, in which coherent light being suitable for optical fiber communication network is used and high security is secured, is provided. A sending end encodes random numbers so that symmetry probability distributions can be obtained at a receiving end, and also sets light intensity and a modulation index of signal light radiating from the sending end so that the SNR of an eavesdropper is less than 2 dB even when said eavesdropper uses a most suitable receiving equipment at the sending end, and also so that the SNR of the receiving end is more than −10 dB, and transmits signals. The receiving end calculates probability distributions of obtained signals and sets a discrimination threshold value after a set of random numbers was transmitted from the sending end. When the probability distributions have some abnormal states, it is judged that the eavesdropper exists, and distributing the cryptographic key is stopped and a fresh cryptographic key is distributed again.

The invention relates to an urban road traffic parameter estimation and road situation discrimination method based on RFID (Radio Frequency Identification) data and an application system and belongs to the technical field of intelligent transportation. The method comprises the steps of preprocessing the RFID data to realize identification of road network running vehicle tracks and obtain traffic parameters such as travel time, road section travel speed, road section flow and road network OD (Origin-Destination) through calculation; according to traffic parameter history data obtained through calculation, performing clustering analysis to traffic states, determining road situation discrimination thresholds and then discriminating the states according to real-time estimated traffic parameters. The urban road traffic parameter estimation and road situation discrimination method based on RFID data and the application system provided by the invention have the advantages that the traffic state discrimination accuracy is above 90 percent, the algorithm efficiency can meet the real-time demand of updating once per minute, the feasibility and the engineering applicability are better and a strong support can be provided for urban traffic state sensing.

The disclosed is a method for synchronization of the running key that is generated from a shared key and that is used for encryption and decryption in communications encrypted with the shared key using a multi-valued signal. In the method for synchronization, the transmitting node transmits a signal that is formed of a multi-valued signal and that has a predetermined fixed pattern before transmitting data encrypted with the shared key. The receiving node generates a bit discrimination threshold signal that allows for bit discrimination and that has a fixed length, shifts bit by bit the phase of the bit discrimination threshold signal while monitoring bit discrimination with respect to a fixed pattern signal that is to be received, and sets the phase of the bit discrimination threshold signal when the phase of the fixed pattern signal matches the phase of the bit discrimination threshold signal. Then, synchronization of the running key is performed in accordance with the phase that has been set between the transmitting node and the receiving node.

The invention discloses an image recognition method and device based on a convolutional neural network. The image recognition method comprises the steps that the quality quantification value of an image to be recognized is extracted based on an image quality prediction branch in the convolutional neural network, wherein the quality quantification value is used to represent the degree of recognition of the image to be recognized; the quality quantification value is compared with a quality discrimination preset threshold; if the quality quantification value is greater than or equal to the quality discrimination preset threshold, the image to be recognized is recognized based on an image feature extracting branch in the convolutional neural network; and if the quality quantification value isless than the quality discrimination threshold, the image to be recognized is not recognized. According to the invention, the problem of low image recognition efficiency in the existing image recognition technology is solved; the quality of the image to be recognized is discriminated; the to-be-recognized image whose quality quantification value is less than the quality discrimination preset threshold is rejected; computing resources required by recognition can be greatly reduced; and the image recognition efficiency and accuracy can be improved.

The invention discloses a dynamic setting method of a differential threshold for online monitoring of transformers. The method includes the following steps: classifying and identifying the historicaldata of a transformer monitored in different operating environments and seasons; preprocessing the classified and identified historical data, and removing abnormal data from the historical data; usinga method based on K-Means clustering to dynamically calculate the discrimination threshold of the transformer based on the historical data with abnormal data removed; and based on a Spark distributedparallel computing framework, constructing a transformer equipment comprehensive warning model of threshold alarm based on the discrimination threshold and trend analysis, so as to differentially monitor the discrimination threshold of the transformer. The calculation time of the dynamic threshold is greatly shortened while the accuracy of alarm is improved.

The invention discloses a deep learning-based visual detection method for packing missing strips. Through the design and construction of a Fast-AnoGAN algorithm model, anomaly detection of cigarette box images is achieved. The Fast-AnoGAN algorithm model includes a generation confrontation network module and an abnormality score calculation module. The generation confrontation network module is combined with a WGAN-GP algorithm model for design and comprises a generator network and a discriminator network. Normal images are used for model training. Feature distribution of normal samples is learned. The abnormality score calculation module is configured to calculate an abnormal score value corresponding to the generated image of the to-be-detected picture input training. Anomaly detection is implemented by determining whether the abnormal score value is greater than a defined abnormality discrimination threshold. The invention has the problems of effectively solving the sample imbalancein the deep learning model training, satisfies the dual requirements of real-time and accuracy of product detection under high-speed production, and has the advantages of adaptability and robustnessto the industrial production environment.

The invention discloses a radar signal sorting method based on a PRI histogram. The method comprises the following steps: establishing the PRI histogram of a radar signal pulse flow; sitting differentjitter ranges for a fixed signal and a dither signal; firstly extracting the fixed signal, and extracting the dither signal, specifically comprising: enabling a box mid-value of each PRI overlappingbox in the PRI histogram to compare with a discrimination threshold: if a value exceeding the discrimination threshold is not existent, zero-clearing the box mid-values of all PRI overlapping boxes inthe PRI histogram; otherwise, using a PRI value corresponding to a peak value exceeding the discrimination threshold as the possible PRI value for performing sequence search; if a sequence is successfully searched, eliminating a detected pulse from the radar signal pulse flow, otherwise, zero-clearing the box mid-values of all PRI overlapping boxes in the PRI histogram, and performing the next order of pulse extraction. The method is capable of improving sorting capacity of the dither signal, and has the better sorting effect to pulse data in a complicated environment.

Provided is a hydraulic shovel (1) constituted by multiple components, comprising: a storage device (20) which stores an operation history of a target component, which is a component included in the multiple components and for which a life span is to be estimated, a discrimination threshold value used for classifying a usage mode of the target component into one of multiple modes on the basis of the operation history, and usage mode-specific estimated life spans which indicate estimated life spans of the target component for the respective usage modes; and an arithmetic and control unit (10) for executing a process of calculating the operation time of the target component for each usage mode on the basis of the discrimination threshold value and the operation history of the target component, and a process of estimating the life span of the target component on the basis of the operation time of the target component for each usage mode calculated in the aforementioned process and the usage mode-specific estimated life spans. This can improve the estimation accuracy of the life span of a component relating to an operating machine.

The invention provides a method and device for nondestructive and rapid prediction of shelf life and freshness of fruits. The method comprises the following steps that spectroscopic data of the fruitswith different pieces of storage time is obtained, and the storage time and environmental data of the fruits are recorded; a fruit storage time model is established by adopting a regression algorithm, and based on a prediction value of the established model and combined with a sample number, a weighted and modified prediction value and a discrimination threshold are calculated, thereby judging the freshness of the fruits. The invention provides a method for formulating the shelf life of the fruits according to data of critical change periods of the fruits. Based on the method, an analysis meter for the freshness of the fruits is developed, and comprises an optical system, a circuit system, a control system and a data storage and processing system; the spectroscopic data of the fruits canbe collected, and the model is called to predict the freshness of the fruits. The method significantly improves the freshness of the fruits and the accuracy of the prediction of the shelf life, can achieve the nondestructive, rapid and precise prediction of the freshness and shelf life of the fruits, and provides a technical reference for the development of the rapid analysis meter for the freshness of the fruits.

The invention discloses a mine microseism and blasting signal identification method based on a waveform oscillation starting trend line slope. The method comprises the steps that first, a linear identification equation is acquired, and a linear identification equation Y = k1 + A*k2 + B is obtained based on N groups of microseism evens and N groups of blasting events, wherein k1 and k2 are used as the parameters of the linear identification equation; second, a discrimination threshold value Yf is calculated; third, an event to be identified is identified based on the linear identification equation and the discrimination threshold value Yf, the waveform oscillation starting trend line slope of the event to be identified is calculated to obtain the k1 and the k2, the k1 and the k2 are substituted into the identification equation to obtain a Y, if the Y is smaller than or equal to the discrimination threshold value Yf, the event to be identified is a microseism event, and otherwise the event to be identified is an blasting event. According to the mine microseism and blasting signal identification method based on the waveform oscillation starting trend line slope, calculated quantity is small, identification accuracy is high, conversion from a time domain to a frequency domain is of no need, cost is low, and implementation is easy.

The invention discloses a method of confronting active deception jamming by a multi-station radar. The method mainly comprises steps: after any node radar is selected to carry out matched filter and target detection on a receiving signal, K+M point targets are obtained, the first node radar serves as a reference radar, time alignment operation is carried out on the K+M point targets and the other N-1 node radars, and an echo amplitude corresponding relation between point targets detected by the N-1 node radars and point targets detected by the reference radar is obtained; echo amplitude of the pth point target on the nth node radar is selected, an amplitude ratio feature vector of the pth point target is obtained, false target determination is set to be a false alarm rate for a real target, a discrimination threshold theta for the N node radars is obtained, the amplitude ratio feature vector omegap of the pth point target is classified to be the pth class, and K+M different classes are obtained, clustering operation is carried out, and a final clustering result is acquired; and finally, a determination threshold value epsilon for false point targets is set, and the false point targets included in the final clustering result are acquired.

The factory-calibrated target discrimination threshold of a radar-based motor vehicle back-up aid is adaptively lowered under specified vehicle operating conditions to provide enhanced target detection sensitivity without causing false target detection. The threshold is initialized to the calibrated value on transition to the reverse range, and is thereafter subject to adaptive adjustment so long as a target is not detected and a high-clutter condition is not in effect. The adaptive adjustment is carried out by low-pass filtering the resultant data of the radar sensor, starting with a factory-calibrated noise baseline, to determine the current noise level, and then calculating the threshold from the noise level. The adaptive adjustment is suspended for an interval after target detection, and the threshold is increased when a high clutter condition is detected based on rate of change in vehicle speed.

The invention discloses a self-adaptive zero velocity update pedestrian navigation method based on a MEMS sensor. By analyzing gait during people multidirectional movement, based on output of the MEMSsensor under different motion modals, a zero velocity update discrimination algorithm combines with strapdown inertial navigation resolving for correcting information radiated with time of speed position resolved by correction resolution; a self-adaptive zero velocity correction model is established according to practical motion characteristic of pedestrians, and zero velocity discrimination threshold can be changed with real time. The method better solve the radiation problem of speed position of different pedestrians under different motion modals under pure inertia resolving, and self-adaptability and reliability of pedestrian navigation positioning during people multidirectional movement is increased.

The invention discloses a method and a circuit for detecting energy segmentation of a specific radioactive nuclide. The method is characterized in that a sodium iodide (NaI) crystal and a photomultiplier are used for forming a detector assembly, a high-energy ray is converted to a weak pulse signal, the weak pulse signal is amplified and shaped by a signal amplification circuit into a signal of a certain amplitude to be supplied to a subsequent circuit, the signal after being amplified and shaped is outputted to an upper discrimination threshold and lower discrimination threshold judgment circuit consisting of two low-bias voltage high-speed comparators, a benchmark source is generated by a high-precision low-temperature-drift integration voltage benchmark source circuit, the comparison voltage is supplied to the comparator after being adjusted by a precise potentiometer, and an adjusting potentiometer adjusts the upper discrimination threshold voltage and the lower discrimination threshold voltage. The method and the circuit have the advantages that the hardware is simple and easy, measurement with certain precise can be completed only by a small-sized detector, no large-sized shielding chamber is needed, the cost of the quantitative detection instrument is greatly reduced, and the quantitative detection instrument of the specific nuclide can be purchased by common consumers.

Provided herein are methods and systems to screen for subjects at risk for retinal-associated eye diseases or to monitor the progression/regression thereof. Generally, the method comprises simultaneously or sequentially displaying to a potentially at-risk individual targets of parallel and nonparallel lines, obtaining an orientation discrimination threshold based on the subject's selections and correlating an increase in the threshold compared to that of a normal control or to a previous threshold measured for the subject as indicative of risk for a retinal-associated eye disease or a progression of the eye disease. Also provided are program storage devices and computer program products effective to store instructions or source code to perform the methods.

What is disclosed is a system and method for determining whether a subject is in atrial fibrillation. A video is received of a region of exposed skin of a subject. The video is acquired of a region where a videoplethysmographic (VPG) signal can be registered by at least one imaging channel of a video imaging device. For each batch of image frames, pixels associated with the region of exposed skin are isolated and processed to obtain a time-series signal. A VPG signal is extracted from the time-series signal. The power spectral density (PSD) is computed across all frequencies within the VPG signal. A pulse harmonic strength (PHS) is calculated for this VPG signal. The pulse harmonic strength is compared to a discrimination threshold, defined herein. A determination is made whether the subject in the video is in atrial fibrillation or in normal sinus rhythm.

The invention belongs to the technical field of information fusion, and relates to a method for association of ESM passive information with radar active information. A linear interpolation method is used firstly, ESM data are aligned to the radar measurement moment; then an association discrimination function is established through 1NN nearest neighbor in a classification mode, and a discrimination threshold is established through a similar Newman-Pearson criterion; then information at different moments is fused through a Dempster-Shafer evidence theory; and finally, an association matrix is determined through hard judgment. It is shown that good performance is achieved after a certain time of accumulation through computer simulation.

The invention belongs to the technical field of nuclear radiation measurement and specifically relates to a method and system for detecting the discrimination threshold of a pulse-amplitude discriminator. The method comprises the following steps of: measuring the pulse amplitude spectrum of a nuclear radiation detector by using a multi-channel pulse amplitude analyzer, determining a point M corresponding to the discrimination threshold in the pulse amplitude spectrum, substituting a cross coordination channel address m' corresponding to the discrimination point M in a multi-channel pulse amplitude analyzer channel address, calculating out the best threshold value A' in coordination with the voltage value transformation formula, and adjusting the threshold value of the discriminator to be the best threshold value A'. According to the invention, the threshold value detection is performed by using the pulse amplitude spectrum; and compared with the traditional detection method in which an integral curve is used, the method provided by the invention has the following advantages of short measurement time and high accuracy.

The invention discloses a self-adaptive real-time true random number generation device based on ultrahigh frequency laser chaos. According to the device, a 3dB optical coupler is arranged at an output end of an ultrahigh frequency chaos laser pulse source; two output ends of the 3dB optical coupler are connected with two input ends of a photoelectric balance detector; and a tunable delay line is arranged on a previous input end. An output end of the photoelectric balance detector is connected with an input end of an amplitude limiting amplifier. According to the device, through utilization of an active mode locking pulse laser, an electronic bottleneck problem is overcome. Through utilization of the photoelectric balance detector and the amplitude limiting amplifier as quantification modules, the problem that a random number rate is low is solved. Through adoption of a self-delay balance detection technology, a true random number is generated self-adaptively. The technical problem that in the prior art, an amplitude discrimination threshold value needs to be set and adjusted continuously is overcome. The device is simple in structure and high in practicability.

The present invention discloses a method for identifying a commercially available orange juice beverage based on a three-dimensional fluorescence spectrum and a robust mahalanobis distance. The method comprises: adopting fluorescence characteristic difference between orange juice components and food additives to establish three-dimensional fluorescence standard spectrum libraries of the freshly squeezed orange juice, the 100% qualification rate orange juice and the orange juice beverage, extracting the spectrum gravity center through a gravity center method to form the 100% orange juice standard feature database, measuring the three-dimensional fluorescence spectrum of a sample requiring identification and extracting the gravity center feature data, and calculating the robust mahalanobis distance to discriminate the sample requiring identification, wherein the discrimination threshold is 2.72. According to the method, the defects of complex process, detection time consuming, high detection cost and the like of the traditional chemical detection method are overcome, the rapid screening and identification on the sample requiring identification can be achieved, the advantages of rapidness, safety, reliability simple operation, low cost and the like are provided, the standardization on the fruit juice beverage industry and protection of consumer interests can be easily achieved, and food safety supervision on the fruit juice beverage can be promoted.

The invention discloses an ultrahigh voltage direct current line fault discrimination method based on frequency band measurement impedance. Voltage and current of two sides of a direct line can be sampled, and voltage and current in specific frequency band can be acquired, and then frequency band measurement impedance of two sides and break variable thereof can be calculated, and then the line fault can be discriminated by comparing the break variable with the set fault discrimination threshold value. The ultrahigh voltage direct current line fault discrimination method is advantageous in that by adopting the ratio between the frequency band voltage and the frequency band current, and the frequency band measurement impedance can be used as the fault criterion, and then the frequency band measurement impedance during the internal fault can be determined by the system parameter and the topology, and cannot be affected by the transient voltage and the current amplitude theoretically, and therefore the problem of the conventional transient harmonic protection being easily affected by the transition resistor and the line attenuation, and being not sensitive to the changes of the parameters such as the line impedance parameter, the line length, and the smoothing reactor, and other parameters can be overcome.