321 results about "Similarity distance" patented technology

A vehicle monitoring apparatus for obtaining a distance image including an image data and three-dimensional distance information and for monitoring surrounding conditions on the basis of the distance image, comprises: means for dividing the distance image into a plurality of blocks composed of picture elements, means for collecting the blocks having a similar distance data and for grouping the blocks into an independent group, means for calculating an area size of the independent group, means for extracting the image data as an object data if the area size of the group is larger than a predetermined value and for discarding the image data as a false data if the area size is smaller than the predetermined value, means for preparing a histogram having a lateral axis representing a deviation amount and a longitudinal axis representing a frequency corresponding to a number of the image data, and means for detecting an object and a distance to the object by deleting the frequency having a smaller number of the image data than a threshold value along the lateral axis on the histogram. Thereby, false data can be removed from the image data and a mismatching can be prevented.

The invention relates to an image retrieval method based on image classification and aims to solve the problem of lower retrieval speed by the conventional method. The method comprises the following steps of: firstly, determining the class number of images in the image classification and a training image set; secondly, extracting the content characteristics of the training image set for classifier training to obtain a classifier; thirdly, inputting an image to be retrieved, extracting the content characteristics of the image to be retrieved as the input of the classifier to obtain a retrieval image set which corresponds to the class, and extracting the content characteristics of each image in the retrieval image set; and finally, according to the obtained content characteristics, acquiring similarity distances of the image to be retrieved and each image in the retrieval image set by using a similarity calculation algorithm, sorting the distances to obtain N images which are nearest to the image to be retrieved, and outputting the N images. The image retrieval method based on the image classification has the advantage that: by combining an image classification technology and the conventional content-based image retrieval method, an image retrieval speed is greatly improved.

The invention provides a text information associating and clustering collecting processing method based on a domain knowledge model. The method comprises the following steps that a text information training set is searched, stemming preprocessing is conducted, and feature word vectors of a text participle sequence of the information training set are extracted through Chinese named entity identification and domain dictionary query modes; representative feature words of a target event are extracted through topic graph model learning training, and a weighted value of topic associating affiliation is calculated; a feature word set is built according to the topic associating affiliation weighted value, calculated through training, of the feature words, and an event topic word template is built; feature word vectors of a participle sequence accessed to text in real time are extracted through the Chinese named entity identification and domain dictionary query modes; the similarity distance of the feature word vectors and all the target event knowledge templates is calculated; the association relationship of multiple texts to the same topic target event is determined according to the similarity threshold, and classification reorganization is conducted by means of a similarity distance ordering rule.

The invention discloses a DCGAN (Deep Convolutional Generative Adversarial Network)-based spectral imagery secure retrieval method, and belongs to the field of spectral imageries. According to the method, the features of a spectral imagery are highly expressed by utilizing a DCGAN; and a new encrypted domain spectral imagery secure retrieval method is proposed. Firstly the deep spectral-spatial features of the spectral imagery are jointly extracted by utilizing the DCGAN, and the contents of the spectral imagery are accurately represented; in order to ensure the security in a remote sensing image retrieval process, the deep features are encrypted by adopting a Min-Hash method based on a criterion that the similarity of the encrypted features is unchanged, thereby protecting the deep features; and finally under the non-decryption condition, Jaccard similarity distance measurement is performed on image features directly by comparing the number of same Min-Hash values, and images similar to a query image are returned. Therefore, the information security is ensured while the retrieval is realized.

The invention discloses a remote sensing image change detection method based on neighbourhood similarity and threshold segmentation and aims to overcome the defect that the traditional method has poor noise immunity and low detection accuracy in terms of the change detection of the target with high noise. The realization process comprises the following steps: (1) using the strength normalization formula to carry out gray level matching on two remote sensing images; (2) using neighbourhood similarity distance measure to construct a similar matrix of the two remote sensing images; (3) combing the similar matrix to construct a difference image of the two remote sensing images; (4) constructing a two-dimension gray level column diagram for the difference image, using the 2D-OTSU method to determine the segmentation threshold value and separating the target area from the background area; and (5) using the fuzzy entropy method to continue classifying the unprocessed edges and noise points. The invention has the advantages of good noise immunity and high detection accuracy for the changing target and can be used for detecting targets with changes of multitemporal remote sensing images.

The invention belongs to the field of multimedia information searching, and particularly discloses a trademark identification searching method for multiple combined contents. The trademark identification searching method for the multiple combined contents aims to overcome the defect that identification result errors are large in the prior art. According to the trademark identification searching method for the multiple combined contents, first, each trademark picture in a model data base is cut to obtain a character part and a figure part, and then characteristic information of two parts are respectively extracted, and characteristic data of all pictures and characters are respectively merged to generate a picture characteristic data base and a character characteristic data base; secondly, trademark image characteristics match with characteristics in characteristic data bases according to a similarity measurement mechanism of images of the multiple combined contents, the similarity distance between a target picture and each trademark model is worked out to obtain a primary identification and searching result, and the result is input to a user; afterwards, second processing is conducted on the primary identification result through a user feedback mechanism, and a final identification and searching result is obtained.

The invention discloses a search method of SAR images classified based on Gauss hybrid model, which mainly solves the problem that the existing SAR image search method has long search time and low precision. The search method comprises the following steps of: establishing SAR image library (I1, I2, ..., Ik), and selecting legible SAR images with relatively even lamellation (I1, I2, ..., Il); extracting the characteristic vectors of all images (f1, f2, ..., fn); classifying the selected SAR images (I1, I2, ..., Il) into (c1, c2, ..., cm), and using the corresponding characteristic vectors as training samples to train the Gauss hybrid model; using the trained Gauss hybrid model to classify the whole image library (I1, I2, ..., Ik) so as to obtain an image library with classification label; extracting a characteristic vector f ' for the inquired image I' input by a user, and using the trained Gauss hybrid model for classification to obtain a classification number ci; and calculating the similarity distances between the inquired image I' and the region comprehensive characteristics of all images of ci classification in the library, and returning the required amount of images of the user according to an ascending distance order. The invention has the advantages of high search speed and high search precision and can be used for searching a large amount of SAR images.

The invention relates to an image feature extraction and similarity measurement method used for three-dimensional city model retrieval. Features extracted through most image and three-dimensional model retrieval methods lack or ignore description of model details, and accordingly, the three-dimensional model retrieval precision is not high. The invention provides a three-dimensional city model retrieval frame based on images. Firstly, retrieval targets on the images are obtained through division, meanwhile, a light field is used for conducting two-dimensional exchanging on three-dimensional city models, features of query targets and features of the retrieval model images are extracted, finally, the similarity between the features is measured through the similarity distance, and three-dimensional city model retrieval is realized. The image feature extraction and similarity measurement method has the advantages that the three-layer frame for image feature extraction and similarity measurement is provided, multiple layers of multi-scale convolutional neural network models with spatial constraints are designed in the frame, and the distinguishable features with invariable displacement, scales and deformation are obtained; a novel similarity measurement method is provided, and similarity matching between the targets is better realized. Compared with an existing method, the efficiency and the precision of the method in three-dimensional city model retrieval are greatly improved.

The invention discloses an image retrieval method, and the method comprises the following steps: obtaining a to-be-retrieved target image, and inputting the target image into a target deep learning model; utilizing the target deep learning model to perform feature extraction on the target image to obtain image features of the target image; wherein the image features comprise global features, localfeatures and multi-scale global features, and the multi-scale global features are features obtained by performing weighted calculation on a plurality of intermediate-stage features generated in the global feature extraction process; respectively calculating similar distances between the target image and the images in the image library by utilizing the image features according to a distance calculation rule; and determining and outputting a similar image of the target image by using the similar distance. According to the method, the image retrieval accuracy can be improved. The invention further discloses an image retrieval device and equipment and a readable storage medium which have corresponding technical effects.

The invention discloses a generation method for image convolution characteristics based on the top layer weight. The generation method comprises steps of downloading images from the Internet and forming an image training set; training a model of a convolutional neural network; using the trained model of the convolutional neural network to extract depth convolution characteristics of different layers of each image respectively; using the obtained depth convolution characteristics to calculate a convolution weight image of the top layer; exerting the effect of the convolution weight image of the top layer on convolution characteristics of from a shallow layer to a high layer to obtain new convolution characteristics; obtaining depth characteristics added with the convolution full value of each image; and by extracting characteristics of the top layer convolution weight of a query image data set and an evaluation image data set respectively, calculating the similarity distance, performing the final similarity coupling, and obtaining a final retrieving result. Compared with prior art, the generation method is suitable for goods in the middle area and goods in any positions. The new top layer weight characteristics are more effective and accurate than the previous gauss weight, and the robustness and the accuracy of the image characteristics can be ensured.

The invention discloses a multi-measurement time series similarity analysis method applicable to k-neighbor inquires of a time series. A multi-single-similarity-measurement method is chosen according to the analysis requirement, each single similarity measurement is used to analyze and inquire an m-neighbor sequence or subsequence of the sequence, pruning the m-neighbor sequence or subsequence under each similarity measurement to obtain a candidate similarity sequence or subsequence, and combining the candidate similarity sequence or subsequence by using a multiple-classifier combination method with advantage weight to obtain the k-neighbor sequence of the inquired sequence. Compared with the single similarity measurement, the similarity analysis of combined multiple measurements can obtain a more comprehensive analysis result. The multiple-classifier combination method with advantage weight regulates the ranking score according to the difference of the similarity distance between the adjacent candidate similarity sequence or subsequence and the inquired sequence while using a BORDA counting method for reference, so as to reflect the specific difference of similarity of the candidate similarity sequence or subsequence.

The invention discloses an intelligent graphic retrieval method, which is characterized in that: extracting features of graphics to generate feature set by the method of Fourier change, training RBF neural network classification model by taking one part of the feature set as training set, indexing the graphics using the classification result given by the classification model; client of a retrieval system extracts the features of retrieval graphics, gives a category by the trained classification model and computes the similarity distance between the retrieval graphics and each graphics in the feature set of the same category; sorting the similarity distance, returning to the graphics according to the number made by the system and further revising RBF neural network classification model by relevant feedback methods. The invention improves the intelligence of search process, effectively determines the RBF neural network classification model by improved algorithm of subtractive clustering, greatly improves retrieval precision, speeds up retrieval speed and upgrades retrieval performance.

Methods, systems and computer program product embodiments for hashing techniques for determining similarity between data sets are described herein. A method embodiment includes, initializing a random number generator with a weighted min-hash value as a seed, wherein the weighted min-hash value approximates a similarity distance between data sets. A number of bits in the weighted min-hash value is determined by uniformly sampling an integer bit value using the random number generator. A system embodiment includes a repository configured to store a plurality of data sets and a hash generator configured to generate weighted min-hash values from the data sets. The system further includes a similarity determiner configured to determine a similarity between the data sets.

The invention discloses a short-term traffic flow forecasting method based on three-layer K nearest neighbors. The short-term traffic flow forecasting method comprises the steps of: (1) counting traffic flow based on fixed time intervals and establishing a historical sample database; (2) evaluating shape similarity between a current point and points in the historical sample database by adopting similarity deviation degree and correlation coefficient respectively, and performing first-layer screening of points; (3) evaluating the points screened in the first layer according to hit rate and shape similar distance, and performing second-layer screening of points; (4) and evaluating matching distances between the current point and the points screened in the second layer by using an Euclidean distance method, and outputs a forecasting result by adopting a weighted mean value of inverse similar distance of a combination shape of the traffic flow at the corresponding next moment when nearest neighbor points are translated to the current point. The short-term traffic flow forecasting method adopts a two-layer shape similarity matching function, takes the shape matching distances between the nearest neighbor points and the current point into account, and improves accuracy and timeliness of short-term traffic flow forecasting.

The invention discloses a method for predicting the complete residual life of an aero-engine under variable working conditions based on working condition identification and similarity matching. According to the method, the problem that the complete residual life is difficult to predict due to degradation trend of the real performance, covered by complex working condition variation, of the aero-engine is focused to be solved. The method comprises the following steps: identifying operation conditions of the aero-engine; carrying out data standardization on historical degradation data under different working conditions; carrying out sensor selection and parameter dimensionality reduction on the standardized data; matching degeneration tracks of each reference engine and a to-be-predicted engine by virtue of a similarity matching method, so as to obtain an estimated value of the residual life of the to-be-predicted engine and the similarity distance between the reference engine and the to-be-predicted engine; and generating a weight according to the similarity distance, and weighing life estimated values, so as to obtain the residual life of the to-be-predicted engine. By verification, the method has relative high prediction accuracies for different engine test samples.

The invention discloses a knowledge learning and privacy protection based big-data user purchase intention predicating method which comprises following steps of: (1) performing normalization processing on a large number of historical data and a small number of current data; (2) grouping the data and establishing a training sample set; (3) counting user purchase intention probability of each group; (4) calculating group labels; (5) training the training set by using an improved support vector machine; (6) constructing a prediction function; (7) inputting to-be-predicted data into the predication function to obtain a prediction result. As the improved support vector machine is used in the method, the small number of current data set probability information and the large number of historical data set probability information are blended into a structural risk minimization learning framework, learning of knowledge in different periods is realized by virtue of constructing similar distance items among data, and accordingly, the knowledge learning and privacy protection based big-data user purchase intention predicating method which is applicable for learning problems of big samples is constructed.

The present application discloses a vehicle searching method and device, which can perform the steps of: calculating an appearance similarity distance between a first image of a target vehicle and several second images containing the searched vehicle; selecting several images from the several second images as several third images; obtaining corresponding license plate features of license plate areas in the first image and each of the third images with a preset Siamese neural network model; calculating a license plate feature similarity distance between the first image and each of the third images according to license plate feature; calculating a visual similarity distance between the first image and each of the third images according to the appearance similarity distance and the license plate feature similarity distance; obtaining a the first search result of the target vehicle by arranging the several third images in an ascending order of the visual similarity distances. The solution provided by the present application is not limited by application scenes, and it also improves vehicle searching speed and accuracy while reducing requirements of hardware such as cameras that collect images of a vehicle and auxiliary devices.

Provided is a data clustering method for rapidly determining a clustering center, comprising the steps of: 1) reading an original data set, selecting a corresponding distance calculating method through dominance analysis, and solving the distance matrix of a whole data set; 2) rapidly determining a clustering center; and 3) selecting optimal dc: 3.1, finding the maximum value dmax and the minimum value dmin in a similarity distance matrix, and calculating a current dc value through setting a percent value; 3.2, after dc is selected and a clustering result is obtained, designing a Fitness function as an evaluation index; 3.3, employing a climbing algorithm to select optimal dc; and 3.4, outputting the optimal dc and the clustering result of the optimal dc. The data clustering method possesses the characteristics of higher accuracy, and smaller difference and parameter dependency of different data set clustering effects.

The invention relates to a selection method for near infrared spectrum modeling samples. According to the selection method for the near infrared spectrum modeling samples, during a blending process, when near infrared spectrums of new gasoline samples are obtained, a plurality of similar samples are selected from an existing gasoline sample spectrum bank through similarity distance to construct an initial training set; performing dimension reduction on the above training set through a principal component analysis method, confirming the optimal principle components and then removing singular points through a T2 statistics method; adopting an appropriate modeling method after confirmation of the training sample set to construct a near infrared spectrum model to perform quantitative analysis on the gasoline samples. The selection method for the near infrared spectrum modeling samples has the advantages of effectively reducing the singular points simply caused by similar distance selection and improving the accuracy and the robustness of the near infrared spectrum model due to the fact that the Hotelling T2 statistics are introduced into the training sample set selection, providing evidence for online modeling and being benefited to change the current situation that the near infrared model during the gasoline blending process cannot be updated in real time.

The invention discloses a method for the matching pursuit of a pedestrian target under an illumination environment change condition. The method is implemented specifically by the following steps of: determining the characteristic range of the target; extracting a reflection component of a target area on the basis of a Retinex principle; performing color transfer correction on the target area; performing characteristic extraction on a target reference template and a target to be matched; calculating a characteristic resemblance distance between the target to be matched and a target reference; and finally performing matching pursuit judgment on the pedestrian target. According to the method, the Retinex principle is utilized, and the color transfer correction is realized, so that the influence of an illumination environment change on the color characteristics of the target is lowered; the pedestrian target is divided into an upper half body and a lower half body, and the color characteristics of the pedestrian target are calculated, so that the influence of the posture change of the pedestrian target when the pedestrian target walks on the color characteristics of the target is lowered; and by the weighted fusion matching of a plurality of characteristics, the accuracy of the matching pursuit of the pedestrian target in an illumination intensity change environment is improved.

A BERT-based customer service question and answer system belongs to the technical field of data calculation and comprises a receiving module, a preprocessing module, an intention module and a template engine module. The receiving module is used for receiving questions proposed by a user side; the preprocessing module is used for processing the received problem; the intention module is used for analyzing and acquiring the intention of the acquired problem; the template engine module is used for matching the obtained questions with standard questions to obtain question methods; an answer configuration module is used for generating answers for the questions provided by the system. According to the system, a BERT model is adopted for feature vector extraction; monitoring is carried out based on a triplet loss function of the Euclidean distance; compared with the adoption of a dichotomy cross entropy loss function, the generated vectors are more natural and reasonable in similarity distance calculation, and compared with a conventional training model, the triplet net simultaneously trains positive and negative samples, so that the model convergence is faster; meanwhile, the data in the system is in a closed-loop state, the modification period is shortened, and the accuracy of the system is improved.

The invention relates to a fabric flaw detection method. Firstly, a nonlinear gray scale co-occurrence matrix characteristic is constructed in order to utilize characteristic space fully and extract image characteristics effectively, then an optimal scale direction parameter and a self-adaptation flaw segmentation threshold of nonlinear gray scale co-occurrence matrix characteristic extraction are obtained through learning of a flawless fabric image, finally, the obtained parameter is employed to characteristics of an image to be detected, and a flaw area is positioned through characteristic similarity distance measurement. The method can position a fabric flaw area effectively, and noise interference is small.

The invention discloses a quick image retrieval method based on reference image indexes, comprising two parts: index and retrieval, wherein the index part comprises the following steps of: extracting the characteristics of images in an image library, randomly selecting an image from the image library as a reference image, and calculating the similarity distance between each image and the reference image by a characteristic similarity comparison method, thereby calculating index numbers; and sorting the index numbers to form an index sequence; and the retrieval part comprises the following steps of: extracting the characteristic of an inquired image; calculating the similarity distance between the inquired image and the reference image to calculate the index number; and obtaining the neighbors of the inquired image in the index sequence by a binary search method, thereby obtaining a similar image set. The method forms mapping of the image characteristics to the one-dimensional real number axis by the characteristic similarity comparison method by introducing the reference image, establishes the indexes and effectively reduces the number of images needing to ba accessed, thereby accelerating the image retrieval speed.

The invention provides a document similarity distinguishing method based on Fourier transform. The method comprises the following steps: acquiring the keyword sequence Ks of a document collection S and a corresponding keyword frequency collection Ns, as well as a keyword sequence Ks' of the detection document s' relative to the document collection S and s corresponding keyword frequency collection Ns'; calculating the weight coefficient of each of the keyword sequences Ks and Ks' as well as the weight sequence FKs of the keyword sequence Ks and the weight sequence FKs' of the keyword sequence Ks'; carrying out Fourier transform to weight sequence FKs and FKs'; calculating the threshold value Omega S of similarity distance of similarity of random document in the detection document s' and the document collection S; calculating the similarity distance D (s', si) between the documents si in the detection document s' and the document collection S, and comparing the similarity distance D with the threshold value Omega S; judging whether the detection document s' and the document collection S are similar or not. The distinguishing method of document similarity based on Fourier transform provided by the invention can not only reduce the requirement to a representing method of the document while calculating similarity, but also can reduce the complexity of calculation and improve the computational efficiency.

The invention discloses an image retrieval method and apparatus. The image retrieval method comprises the following steps of building a multi-task depth network structure model; establishing a target image feature library; inputting a feature subset of a to-be-retrieved image; calculating a similarity distance between the feature subset of the to-be-retrieved image and each image feature in the image feature library; and performing arrangement according to an arrangement order of distances from short to long to obtain an image with a minimum distance with the to-be-retrieved image. By building the depth network structure model for multi-task learning of the features of the target image, the image retrieval is finished; the precision of a plurality of attribute classifications can be jointly improved by fully utilizing a correlation of tasks; and a relationship of the attributes of the image and detail features of the image can be learnt, so that in the image retrieval process, the influence of climate, environment, illuminance and the like can be overcome.

The invention discloses a test-case selection method based on user sessions and a hierarchical clustering algorithm. The method includes the following steps: acquiring server access logs, and carryingout sorting according to time; carrying out preprocessing and clustering to form a user session sequence set; calculating similarity distances among all user session sequences through using an improved user-session-sequence comparison algorithm; employing the improved condensing hierarchical clustering algorithm to cluster the user session sequences, and outputting final clustering results of test cases; and optimizing selection of the test cases through deleting redundant test cases. According to the method of the invention, representative user operation sequences can be quickly mined from the large number of server access logs to use the same as test cases, automation of test-case generation and optimization of test-case selection are realized, and subsequent work of automated functiontests of a server, performance tests, user behavior analysis and the like is facilitated.

The invention discloses a remote sensing image content retrieval method for semi-supervised deep adversarial self-coding Hash learning. The method comprises the steps of establishing a remote sensingimage feature library, and selecting a plurality of samples as training samples; training an adversarial self-coding Hash learning model by using the training sample; performing Hash coding on the whole remote sensing image feature library by using an adversarial self-coding Hash coding model to obtain a Hash database; processing a query image input by a user, obtaining a feature vector corresponding to the query image through the same pre-training network, and performing Hash coding by using a confrontation self-coding Hash learning model to obtain a corresponding Hash code; and finally, calculating similar distances between the query image and all images in the image library, returning the images required by the user according to the distances from small to large, and finding the corresponding image in the remote sensing image library according to the index to complete image retrieval. According to the method, high retrieval precision can be kept under semi-supervised learning, Hashcoding is more efficient, smaller quantization loss is achieved, and the retrieval precision is further improved.

The invention relates to a method for detecting uniqueness of a person entering a coal mineral well. The method comprises the following steps: receiving and measuring a RSSI (received signal strength indicator) of a label by a base station; establishing a RSSI sequence database; according to label sending data, periodically selecting a slide window, and filtering the RSSI in the window; fitting the filtered RSSI to be a curve line; freely selecting two curve lines after filtering and fitting, and performing vertical translational treatment on one curve line; apply a dynamic time wrapping algorithm and obtaining a similarity distance corresponding to an optimal path between two curve lines; comparing the obtained minimum similarity distance with the set threshold value, exactly judging if it is one person with multiple blocks. According to the method, the label RSSI similarity is compared on the basis of the dynamic time wrapping algorithm, and further the space position of the label is judged; the method realizes the uniqueness detection of the person who enters the coal mineral well, and solves the problem the current label-based mining person location system cannot accurately detect the uniqueness of the person who enters the coal mineral well.

A method for implementing automated information technology (IT) system failure recommendation and mitigation includes performing log pattern learning to automatically generate sparse time series for each log pattern for a set of classification logs corresponding to a failure, performing multivariate log time series extraction based on the log pattern learning to generate a failure signature for the set of classification logs, including representing the sparse time series as a run-length encoded sequence for efficient storage and computation, calculating a similarity distance between the failure signature for the set of classification logs and each failure signature from a failure signature model file, determining a failure label for the failure corresponding to a most similar known failure based on the similarity distance, and initiating failure mitigation based on the failure label.

The invention discloses a face comparison method and device, computer equipment and a storage medium. The method comprises the steps: training a first comparison model based on a preset color image through a deep neural network, and building a twin network according to the first comparison model; respectively inputting the positive sample and the negative sample into a twin network to obtain feature vectors of the user color image and the user infrared image in the same sample; calculating a cosine similarity distance and KL divergence between the feature vectors of the two images, substituting the cosine similarity distance into a preset comparison loss function to obtain a comparison loss value, training a twin network according to the comparison loss value and the KL divergence, and obtaining a second comparison model after sample data iteration; inputting the target color image and the target infrared image into a second comparison model, and calculating and outputting a face comparison result according to the similarity, thereby solving a problem that there is a recognition difference between the color image and the infrared image in face comparison.