673 results about "Mahalanobis distance" patented technology

An image forming apparatus is capable of specifying the type of an occurring abnormality to a certain extent while avoiding the increasingly complicated control that is caused when a plurality of abnormalities is detected individually according to the presence of their respective causes. A normal group data set, which is a collection of normal value combinations relating to grouped information constituted by a plurality of information of differing types, is stored in RAM or ROM serving as information storage means of a control unit. At least two or more sets of grouped information, comprising first grouped information constituted by a plurality of different types of information, and second grouped information constituted by a plurality of information in a different combination to that of the first grouped information, are obtained from the RAM, the ROM, various sensors, and an operation display unit. A CPU serving as determining means determines a Mahalanobis distance for each set of grouped information on the basis of the normal group data set and the obtained results of each set of grouped information, and uses this Mahalanobis distance to determine the presence of an abnormality according to categories.

A failure diagnosis apparatus for a refrigerating cycle had a problem that it has a low precision because the fluid is treated, and it is difficult to detect a foretaste of failure, absorb individual differences of real machine in the failure determination, and determine a cause of failure. Also, no cheap and practical diagnosis apparatus and method are provided. A plurality of instrumentation amounts concerning the refrigerant such as the pressure and temperature of the refrigerating cycle apparatus or other instrumentation amounts are detected, the state quantities such as composite variables are acquired by making the arithmetic operation on these instrumentation amounts, and whether the apparatus is normal or abnormal is judged employing the arithmetic operation results. If learning is made during the normal operation, a current state is judged, and if learning is made by compulsorily performing the abnormal operation, or if the abnormal operating condition is operated during the current operation, a failure foretaste such as a critical operation can be made from a change in the Mahalanobis distance. Thereby, the secure diagnosis can be implemented with a simple constitution.

A computer system implements a risk model for detecting outliers in a large plurality of transaction data, which can encompass millions or billions of transactions in some instances. The computing system comprises a non-transitory computer readable storage medium storing program instructions for execution by a computer processor in order to cause the computing system to receive first features for an entity in the transaction data, receive second features for a benchmark set, the second features corresponding with the first features, determine an outlier value of the entity based on a Mahalanobis distance from the first features to a benchmark value representing an average for the second features. The output of the risk model can be used to prioritize review by a human data analyst. The data analyst's review of the underlying data can be used to improve the model.

The invention discloses a semantic annotation method for a hyperspectral remote sensing image. The semantic annotation method comprises the following steps of: I, acquiring training data and test data of the hyperspectral remote sensing image through spectral information and an annotated truth value of the hyperspectral remote sensing image; II, constructing a convolutional neural network according to the number of bands of the hyperspectral remote sensing image; III, training the convolutional neural network through the training data to obtain a convolutional neural network model; IV, classifying the test data through the convolutional neural network model to obtain a semantic annotation result; V, constructing a unary potential-energy function of a conditional random field model according to the semantic annotation result; VI, constructing a binary potential-energy function of the conditional random field model in a neighborhood by using an edge constraint model based on an improved mahalanobis distance; VII, carrying out weight adjustment of the unary potential-energy function and the binary potential-energy function on the conditional random field model; VIII, solving the conditional random field model to obtain the semantic annotation result. Through the above steps, the semantic annotation method for the hyperspectral remote sensing image is realized.

The invention provides a hardware Trojan horse detection method and a hardware Trojan horse detection system. The method includes the following steps: collecting a by-pass signal of a to-be-detected integrated circuit; extracting features of the by-pass signal to form a feature collection; calculating mahalanobis distance values, including a mahalanobis value of a reference integrated circuit and a mahalanobis value of the to-be-detected integrated circuit, of the feature collection; comparing the mahalanobis value of the reference integrated circuit with that of the to-be-detected integrated circuit, and performing hardware Trojan horse detection according to a comparison result. By the hardware Trojan horse detection method and the hardware Trojan horse detection system which have the advantages of simple algorithm and short detection time, distinguishability and efficiency of hardware Trojan horse detection in integrated circuit testing are effectively improved, and any cost for hardware is not generated.

A computer-implemented method to provide a desired variable subset. The method may include obtaining a set of data records corresponding a plurality of variables and defining the data records as normal data or abnormal data based on predetermined criteria. The method may also include initializing a genetic algorithm with a subset of variables from the plurality of variables and calculating Mahalanobis distances of the normal data and the abnormal data based on the subset of variables. Further, the method may include identifying a desired subset of the plurality of variables by performing the genetic algorithm based on the Mahalanobis distances.

The invention discloses a personalized parking space recommendation method and a system. The method comprises the following steps: collecting an original data, and a minimum driving time Tk to reach the parking space, a parking fee F, a parking success probability C, a walking distance M from the parking space to the destination, a parking difficulty D to the parking space, and a parking space safety S which are calculated from the original data; performing cluster analysis by utilizing a K-MEDOIDS algorithm; selecting and matching the center point of each cluster after clustering with the preference weight value of a predetermined parking space; and recommending the parking space suitable for the driver to a specific driver. According to the personalized parking space recommendation method and the system, user-defined weight value and historical data adjustment weight value are introduced by fully considering the characteristic that different drivers have different parking needs; theaggregated cluster is higher in quality, and is closer to the demand of users by improving the Mahalanobis distance formula through the weight values.

The invention discloses an online detection method and protection device for direct current system arc faults. Characteristics of current alternating components of a power circuit in the power-on starting process, the normal work process and the overload fault process of the power circuit and characteristics of the current alternating components of the power circuit during arc generating are combined, selected arc current characteristic components include peak-to-peak values and standard deviations of a time domain and a frequency component power sum within the range from 1 kHz to 100 kHz after noise of specific frequency points of a load is filtered, mahalanobis distance values of the real-time characteristic components and the characteristic components in the power-on starting process, the normal work process and the overload fault process of the power circuit through a mahalanobis distance algorithm, and the mahalanobis distance values are compared with a threshold value set in the learning process to detect whether a single arc is generated or not. By means of the technical scheme, the online detection method and protection device have advantages of being high in university, high in detection rate and low in misjudgment rate, and can be widely applied to low-voltage and high-voltage electric power systems and photovoltaic cell systems of airplanes, electric vehicles and ships, and civil high-voltage direct current power distribution systems.

The invention discloses a rolling bearing failure diagnosis method base on vibration temporal frequency analysis. The method comprises the following steps: utilizing a vibration acceleration sensor to collect vibration signals of the rolling bearing under a normal condition and a failure condition; utilizing a modified inherent time scale resolving method to resolve the collected vibration signals, and generating a plurality of inherent time scale components and residual signals; calculating relativity of the time scale components and the vibration signals, selecting the inherent time scale components of which the relativity is ranked top 5 as related components, and rejecting noise signals and false components; calculating Wigner distribution of the related components respectively, and conducting linear stack to obtain the Wigner temporal frequency figure of the original signal; extracting difference fractal box dimensionality of the Wigner temporal frequency figure and the image entropy as failure characteristics; utilizing mahalanobis distance to build mapping relation of the failure characteristics and failure types to realize failure diagnosis. According to the invention, interference of Wigner distribution cross terms is avoided; two kinds of representative failure characteristics of the difference fractal box dimensionality and the image entropy are confirmed.

The invention discloses a pedestrian re-identifying method based on coordination scale learning and belongs to the technical field of monitoring video retrieval. First, according to color and texture features of images in a marked training sample set L, scale learning is carried out, and covariance matrixes Mc and Mt in corresponding Mahalanobis distance are obtained; and checking targets are selected randomly, the Mc and the Mt are used for Mahalanobis distance measuring, a corresponding sorting result is obtained, positive samples and negative samples are obtained, a new marked training sample set L is obtained, the Mc and the Mt are updated until an unmarked training sample set U is empty, a final marked sample set L* is obtained, the color and texture features are fused, an Mf is obtained, and a Mahalanobis distance function based on the Mf can be used for pedestrian re-identifying. Under a semi-supervised framework, the pedestrian re-identifying technology based on scale learning is studied, scale learning is carried out with the marked samples assisted by the unmarked samples, the requirement that practical video investigation application marked training samples are hard to obtain is met, and re-identifying performance under few marked samples can be effectively improved.

The invention discloses a method of assisting cigarette formula by adopting an SIMCA (Soft Independent Modeling of Class Analogy) based on Near-infrared spectral information, comprising the following steps of: (1) modeling sample preparation; (2) spectrum scanning; (3) spectrum pretreatment; (4) principal component analysis; (5) tobacco raw material database establishment; (6) substitution rule setting; and (7) formula assisting: during the tobacco raw material substitution, scanning the to-be-detected sample according to the steps 1-4 by taking a to-be-substituted tobacco sample as a target, obtaining the near-infrared spectral data after treating, setting the substitution rule on the basis of the information of the substituted sample according to the step 6, comparing the near-infrared data of the substituted sample with a mathematical model in a memory to obtain replaceable tobacco samples, sequencing the samples according to mahalanobis distance, wherein the sample with smaller mahalanobis distance is more similar, and finally carrying out sensory taste. The method disclosed by the invention is reliable and capable of reducing the range of looking for substitute samples for formula personnel, greatly reducing the workload and enhancing the pertinence of tobacco formula work.

The invention discloses a health assessment and fault diagnosis method for rotating machinery based on fisher discriminant analysis and a mahalanobis distance, which belongs to the technical field of condition-based maintenance of the rotating machinery. The method comprises the steps of extracting an energy eigenvector based on wavelet packet decomposition, constructing a discriminant analysis function, conducting health status assessment, conducting fault detection on the rotating machinery, and finally conducting fault diagnosis on the rotating machinery. The method constructs a comprehensive frame integrating the status assessment, the fault detection and the fault diagnosis, solves the hotspot problem in comprehensive health management of the rotating machinery at present, achieves intelligent maintenance of the rotating machinery, can establish an assessment and diagnosis model without full life status monitoring data of the rotating machinery, reduces the dependence on historical data, and is very high in engineering applicability.

The invention provides a KNN-based improved missing data filling algorithm, which comprises the steps of (1) improving a traditional multiple correlation coefficient inverse weighting method and calculating the importance of each attribute on a missing value-containing attribute by using an improved algorithm, deleting a few of attributes with relatively small correlation with a key attribute and carrying out streamlined operation on an attribute set to obtain a data sample set which only contains the streamlined attribute set; (2) comprehensively considering the advantages of the correlation between the attributes and the variability by using a mahalanobis distance, effectively predicting an uncertain factor-containing sample by combining a grey correlation analysis method and calculating K adjacent samples of a missing sample; and (3) giving entropy weight values to the attributes corresponding to the K samples according to the calculated K distance values and an entropy weight method and then calculating a final filling value by combining attribute values. According to the KNN-based improved missing data filling algorithm, the calculating complexity of the missing data algorithm can be reduced, the accuracy of the adjacent sample values is improved and the estimation accuracy of the data filing value is improved.

The invention discloses an automatic image annotation method based on deep learning and canonical correlation analysis. The method includes: using a depth Boltzmann machine to extract the high-level feature vectors of images and annotation words, selecting multiple Bernoulli distribution to fit annotation word samples, and selecting Gaussian distribution to fit image features; performing canonical correlation analysis on the high-level features of the images and the annotation words; calculating the Mahalanobis distance between to-be-annotated images and training set images in canonical variable space, and performing weighted calculation according to the distance to obtain high-level annotation word features; generating image annotation words through mean field estimation. The depth Boltzmann machine comprises I-DBM and T-DBM which are respectively used for extracting the high-level feature vectors of the images and the annotation words. Each of the I-DBM and the T-DBM sequentially comprises a visible layer, a first hidden unit layer and a second hidden unit layer from bottom to top. By the method, the problem of 'semantic gap' during image semantic annotation can be solved effectively, and annotation accuracy is increased.

A method and a system is provided for discriminating between healthy cervical tissue and pathologic cervical tissue based on the fluorescence response of the tissue to laser excitation (LIF) and the backscatter response to illumination by white light (in the spectral range of 360 to 750 nm). Combining LIF and white light responses, as well as evaluating a spatial correlation between proximate cervical tissue sites in conjunction with a statistically significant “distance” algorithm, such as the Mahalanobis distance between data sets, can improve the discrimination between normal and abnormal tissue. The results may be displayed in the form of a map of the cervix representing the suspected pathology.

The invention provides a mechanical wearing part performance assessment and prediction method based on based on EMD (empirical mode decomposition)-SVD (singular value decomposition) and an MTS (Mahalanobis-Taguchi system), and belongs to the technical field of mechanical wearing part fault diagnosis. The method comprises: first of all, performing noise reduction processing on acquired signals of a monitored object, then performing EMD on the signals, selecting effective IMF (intrinsic mode function) components and residual functions to form an initial matrix, performing SVD on the initial matrix, and performing normalization processing on obtained characteristic values to obtain characteristic vectors; then using an MTS method to calculate an MD (Mahalanobis Distance), and using a Taguchi method to perform optimization and reduction on the characteristic vectors; and converting the MD into a confidence value, performing assessment on the performance of mechanical wearing parts through tracking the trend of the confidence value, and performing prediction on a fault through a correlation module or a matching matrix between the confidence value and conditions of the monitored object. The method provided by the invention avoids the problem of easily occurring errors when a conventional method is used for processing non-linear non-stationary signals, and reduces fault generation probability, thereby being suitable for industrial real-time monitoring.

The invention relates to the technical field of computers, and provides a target tracking method and device, equipment and a storage medium. The method comprises the steps: calculating the Mahalanobisdistance between each obtained original detection region and each obtained prediction target region; taking the original detection target corresponding to the mahalanobis distance with the minimum value as a basic target to be evaluated; if the number of the to-be-evaluated basic targets is more than one, taking an original detection target in the previous frame of video image corresponding to the prediction tracking target as a to-be-processed target; performing depth feature extraction on each to-be-processed target to obtain a depth feature vector of each to-be-processed target; and matching the depth feature vector of each to-be-processed target in each frame of video image with the depth feature vector of each to-be-processed target in the next frame of video image to obtain a matching result corresponding to each to-be-processed target in each frame of video image, and completing target tracking according to the matching result. The accuracy of target tracking can be improved.

The invention discloses a power MOSFET health state assessment and residual life prediction method. The concrete steps are that: drain and source electrode voltage, drain and source electrode current and threshold voltage of a health power MOSFET are firstly acquired, mahalanobis distance of the health power MOSFET is acquired and Box-Cox conversion is performed so that the mahalanobis distance in normal distribution is acquired, then a health state assessment reference threshold of the power MOSFET is confirmed, and the mahalanobis distance in normal distribution is selected as a health state characteristic parameter for assessing the measured power MOSFET; then the health state assessment is performed on the drain and source electrode voltage, the drain and source electrode current and the threshold voltage monitoring the measured power MOSFET; and finally different residual life prediction models are established according to different health states of the measured MOSFET. Multi-characteristic parameters of the power MOSFET are converted into a single-characteristic parameter to perform the health state assessment. Meanwhile, temperature and voltage stress of working conditions are considered, and the residual life prediction models under the normal states and the abnormal states are established so that the residual life of the measured power MOSFET can be accurately predicted.

The invention relates to an automatic image semantic annotation method based on scale learning and correlated label dissemination, which comprises the following steps: firstly, the global and partial feature descriptor of each image is extracted after the image library is read; the feature descriptor is sent to a model based on a structured support vector machine for learning the distance scale between the images, actually the Mahalanobis distance; a model about the internal relation between key words is built; the learned Mahalanobis distance is embedded in a built label dissemination model so as to obtain the confidence degree score of each key word belonging to the image to be labeled; and a threshold value is set for the confidence degree score of each key word, and the key words of which the scores are higher than the threshold value are distributed to the images to be labeled, thereby completing labeling. The learning algorithm model based on the structured support vector machine can effectively solve the measuring problem of similarity between the images, the internal relation between the key words is fully excavated through the embedded-type correlated label dissemination model, and the accuracy of the image annotation and image retrieval is effectively improved.

The invention provides an electric motor health monitoring and abnormity diagnostic method based on feature selection and the mahalanobis distance. The method includes the first step of conducting data acquisition on vibration signals, current signals and rotating speed signals of electric motors, conducting feature computing on the signals, constructing feature spaces and selecting feature vectors for calculation of the mahalanobis distance in a feature selection method, the second step of the mahalanobis distances of the electric motors in a normal operation state and constructing mahalanobis spaces indicating the normal operation state of the electric motors, and the third step of calculating the mahalanobis distances according to signals of tested electric motors with unknown health conditions with reference to statistic parameters of the motors in the normal operation state and judging the health condition of the tested electric motors through comparison of the mahalanobis spaces. Through the electric motor health monitoring and abnormity diagnostic method based on feature selection and the mahalanobis distance, since the signals of the electric motors in the normal operation state are used for constructing the mahalanobis spaces, health monitoring and abnormity diagnosis on the motors in an unknown operation state can be effectively achieved.

The invention provides a method for judging the category of failures caused by electrical contact in a sealed electromagnetic relay. The method comprises the following steps: firstly, running tests on the reliability life of a plurality of sealed electromagnetic relays as the samples to be tested; recording the relation between the contact resistance and other characteristic parameters of each sample to be tested and the operation frequency thereof in the entire test process; forming a data matrix Xn*6 from the six characteristic parameters, such as the contact resistance; carrying out the dimension reduction pre-processing on the multi-dimension characteristic parameter data by using the principal component analysis method; extracting the corresponding data characteristics from the data subjected to the dimension reduction, and sorting the data characteristics by failure mechanisms; and calculating the Mahalanobis distances between the new sample to be tested and various training samples using the distance discriminant analysis method, and comparing the Mahalanobis distances to judge the category of the failure. Instead of opening a housing and achieving the failure analysis by optical microscopes and other instruments, the invention can eliminate the interference factors caused by other inducements and expose and locate the real causes of failures.

In a method of controlling a manufacturing process, a Mahalanobis space of plural manufacturing control parameters is generated on the basis of first sampled data. Then, a Mahalanobis distance from the Mahalanobis space and second sampled data is calculated. A manufacturing process is determined to be under a malfunction operating condition by comparing the Mahalanobis distance and a threshold value.

The invention discloses a color recognition method based on an improved SLIC super-pixel segmentation algorithm. The method comprises the following steps: (1) loading a Lab color mode sample set; (2) acquiring a to-be-recognized target color image, performing filtering and correction preprocessing on the target color image; (3) processing the preprocessed target color image through the adoption of the SLIC super-pixel segmentation algorithm to segment a plurality of super-pixel regions; (4) performing mean value processing on each segmented super-pixel region so that all pixel values in single super-pixel region are the same; (5) comparing one pixel value in the super-pixel region with the color in the sample set loaded in the step (1) through the adoption of the mahalanobis distance, wherein the color corresponding to the minimum value of the mahalanobis distance is the color of the super-pixel region. The method of processing every pixel in the traditional color recognition is changed, the operation processing speed and the recognition precision are greatly improved.

The invention provides a traffic alarm condition level predication method based on distance metric learning. The city traffic police situation level is predicated under the condition of known weather data, time data and environment data, the sorted multidimensional historical data is classified according to the requirements of a traffic police command department, a generalized Mahalanobis distance measure method is used to learn the classified and marked multidimensional historical data, and the weight value of each characteristic attribute for a traffic alarm condition level is obtained by a distance metric learning matrix. The classification contribution degree of a characteristic attribute with a large weight value is large, the similarity of the current multi-dimensional data and historical data is calculated according to an Euclidean distance with a weight value, K historical data which is most similar to the current data is selected to carry alarm condition situation level vote, and an alarm condition level with the highest vote is taken as the predication result of a current traffic alarm condition level. The method has the advantages of effective realization of prediction and high accuracy.