191 results about "Digit recognition" patented technology

The invention relates to an image recognition method, in particular to an automatic meter-reading image recognition method, and aims at solving the technical problem of low accuracy and proneness to digital recognition error. The automatic meter-reading image recognition method is simple and convenient to use, capable of accurately recognizing dial plate images and high in accuracy. The technical scheme includes the following steps: a, collecting images of a dial plate by an image collecting device; b, roughly positioning a digit indicating image; c, accurately positioning a digit indicating area image; d, accurately positioning a pointer area image; e, preprocessing the images; f, performing digit division; g, performing digit recognition; h, recognizing angle of a pointer. The automatic meter-reading image recognition method is widely used for recognition of digits of the dial plate in a remote meter reading system.

In many areas, recent developments have generated very large datasets from which it is desired to extract meaningful relationships between the dataset elements. However, to date, the finding of such relationships using prior art methods has proved extremely difficult especially in the biomedical arts. Methods for local causal learning and Markov blanket discovery are important recent developments in pattern recognition and applied statistics, primarily because they offer a principled solution to the variable/feature selection problem and give insight about local causal structure. The present invention provides a generative method for learning local causal structure around target variables of interest in the form of direct causes/effects and Markov blankets applicable to very large datasets and relatively small samples. The method is readily applicable to real-world data, and the selected feature sets can be used for causal discovery and classification. The generative method GLL-PC can be instantiated in many ways, giving rise to novel method variants. In general, the inventive method transforms a dataset with many variables into either a minimal reduced dataset where all variables are needed for optimal prediction of the response variable or a dataset where all variables are direct causes and direct effects of the response variable. The power of the invention and significant advantages over the prior art were empirically demonstrated with datasets from a diversity of application domains (biology, medicine, economics, ecology, digit recognition, text categorization, and computational biology) and data generated by Bayesian networks.

The invention provides a recognition method for English words and digits in a natural scene image. According to the method, recognition of the English words and the digits in a natural scene is divided into a feature extraction step, a feature focusing step and a feature recognition step, wherein a convolutional neural network is utilized to perform feature extraction on an input image, an attention mechanism is utilized to focus on useful information in a feature sequence, a long and short-time memory network is utilized to recognize feature vectors, therefore, a deep neural network and the attention mechanisms are combined, and a final recognition result can be obtained directly when the image is input into the deep neural network. Through the method, it is not needed to perform winding sliding operation on the input image or recognize characters in a window; and meanwhile, an output string is the final recognition result, so that it is not needed to integrate recognized strings through a merging algorithm.

The invention provides a digit recognition method for identification cards. The digit recognition method comprises the following steps: acquiring a classifier training file; transforming a color image to be identified into a gray level image; accurately positioning digit element positions of each identification card; accurately segmenting digit characters of each identification card; verifying whether a candidate region is a real digit region of each identification card; carrying out digit recognition on the basis of a nearest neighbor algorithm; judging whether the current digit element is the last group of digit element. According to the digit recognition method provided by the invention, positioning, segmentation and identification of digits on the identification cards are carried out by adopting an image processing technology; the recognition precision is relatively high, no special equipment is needed and the cost is low; the digit recognition method has strong applicability to various environments and can be used for reading information of multiple identification cards simultaneously.

The invention relates to amethod and device for identifying the digital image displayed in the mechanical instrument window, suitale for the identification of figure for various mechanical instrument window. The method, based on the working principle of absolute encoder in the field of industrial sensor, applys the mechanical location identifying method to the field of identifying image, carry out the identification for the position of image to reach the identification for the content of image, and especially is suitale for the application in which the one-chip microcomputer is as a processor, with high efficiency, accurate identification and good versatilityin. The device adopts the one-chip microcomputer as processor and is an integer of collecting, processing, identifying and transmitting image, and characterized in that the device comprises an image collection module, one-chip microcomputer processing and identifying module, a communication interface, a power module and transparent sealing structure. The device, having low cost, high efficiency and little volume, can be installed above the various mechanical instrument window with other structure, accordingly the identification of figure displayed in the instrument window can be carried out, as well as the date and the image are transmitted at the same time.

The present invention provides a handwritten digital recognition method and system based on depth learning. A convolutional neural network is trained with a sample set which is formed by constructing a handwritten digital picture with a label, the trained convolutional neural network is saved, and the image to be identified is taken as the input, and according to the output vector, the recognition result is obtained. The handwritten digital image is identified through the convolution neural network, so that the displayed feature extraction is avoided, the picture is directly taken as the network input, and the recognition accuracy is high. The network can be repeatedly used after being trained, the processing efficiency is high and the training time is short.

The invention relates to a floating number plates discriminating method, which is based on the three-road parallel manmade nerve-net. The identify steps are that adopting the video burst mode to catch the video automatically for the collection of video pictures of the moving vehicle, character cutting the part of pictures of the number plates and then being the input signal of the nerve-net; said nerve-net adopts three standard self-adapt vibrate net each of which has its own work, the three nets are the Chinese characters identify net, the English characters identify net and the data identify net, all of which identify the inputted vector signals at one time, and outputs the most similar sort unity respectively, through controlling the property of the Territory value and the identify accurately according to the number plates and filter, at last add a color property of number plates as the output of the discriminating results. Adopting the method of the invention can identify the number plates of vehicle fast and accurately under the largest driving speed of the vehicles.

The invention relates o an identification method for handwritten numbers. The method comprises the following steps of preprocessing number images, wherein the preprocessing step comprises the substepsof graying the number images, carrying out binaryzation on the number images, carrying out image denoising, segmenting character strings, carrying out number normalization, and carrying out number refinement; and setting up a deep convolutional neural network model, configuring neural network parameters, generating training set samples and test set samples, adjusting the parameters, training thenetwork model, and identifying the numbers through the trained network mode. According to the method, through preprocessing of the image, influences of noises on the image can be prominently reduced,and the images with different sizes can be normalized into the images with the same size; through utilization of the deep convolutional neural network, the training set samples and the test set samples are generated, the parameters are adjusted, the network is trained, and the numbers are identified through utilization of the trained network model. The handwritten number identification accuracy and rate can be improved under the condition that the complexity is similar, and the method can be widely applied to the fields such as post office letter sorting and bank check inputting.

The present invention discloses a multi-feature fusion and deep learning network extraction-based handwritten digit recognition method. The method includes the following steps that: the image data of handwritten digits are read, vectorization pre-processing is performed on the data; multi-feature fusion is performed on the processed data by using the principal component analysis (PCA) technique and the histogram of oriented gradients (HOG) technique, so that shallow compound features can be constructed; secondary feature extraction is performed on the data which have been subjected to multi-feature fusion through using a shallow stack auto-encoder (SAE) model, so that a deep learning network can be constructed to perform high-level and deep learning and processing on the shallow compound features; and the Softmax classifier is used to test classification effects. According to the multi-feature fusion and deep learning network extraction-based handwritten digit recognition method of the invention, the multi-feature fusion method is adopted, the PCA technique and the HOG technique are integrated, so that the shallow compound features can be constructed; the SAE model is adopted to perform secondary feature extraction, so that the deep learning network can be constructed, and simpler and more efficient feature samples can be obtained; and the Softmax classifier is used to test the classification effects; and therefore, the recognition accuracy rate of handwritten digits can be increased to 99.2%.

An embodiment of the invention discloses a handwritten digit recognition method and system. During dimensionality reduction of handwritten digits, all image data are represented linearly through K neighbors, weighting coefficients of all image data when the image data are represented linearly through the K neighbors are obtained through an orthogonal matching algorithm, besides, weighting coefficient matrixes are established to perform dimensionality reduction on training image data, and dimensionality reduction is performed on images to be recognized through weighting coefficient vectors and vector data of the K neighbors of the images after dimensionality reduction. According to experiments, by the aid of the handwritten digit recognition method, the recognition rate of handwritten digit recognition is improved.

The invention discloses a handwritten numeral recognition method based on point density weighting online FCM clustering. The method is used for processing the large-scale offline handwritten numeral recognition problem. The method includes the steps that (1), all handwritten numeral image sets are preprocessed; (2), clustering centers are initialized, and data points are made to sequentially enter processing procedures; (3), the membership degree of the current data point and all the clustering centers is calculated; (4), if the membership degree reaches a threshold value, the position of the nearest clustering center is updated; (5), if the membership degree does not reach the threshold value, the current data point is not processed and is temporarily placed in a to-be-processed region; (6), when the to-be-processed region reaches certain standards, data in the to-be-processed region are clustered through a point density weighting FCM algorithm, and the clustering centers are updated; (7), circulation continues until all the data points are processed; (8), the membership degrees of all the data points are calculated through acquired clustering center blocks, the data points are divided into different classes, and data classification is finished through scanning at a time. According to the method, the space complexity and the time complexity can be lowered from the aspect of processing the large-scale handwritten numeral recognition problem.

A brain-computer interface (BCI) designed with a hybrid electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) for letter and number recognition system for people who cannot speak. By this system, the words and numbers a subject thinks of are reflected on a display screen. A speech tool has been designed for these individuals allowing them to express themselves.

The invention discloses a multifunctional digital meter automatic reading method based on a target detection algorithm, and the method comprises the following steps: S1, carrying out the lightweight improvement of a YOLOv4 network model, and obtaining a YOLOv4 lightweight network model; S2, collecting the multifunctional digital table picture, and performing position calibration and category calibration on numbers on the multifunctional digital table picture; and S3, training the YOLOv4 lightweight network model by using the multifunctional digital table picture calibrated in the step S2 and the calibration information of the multifunctional digital table picture to obtain a YOLOv4 lightweight digital detection model. The multifunctional digital meter automatic reading method based on thetarget detection algorithm is high in recognition speed, high in recognition accuracy and good in stability, and the problems that an existing digital meter automatic reading method is prone to beingaffected by external environment changes and image imaging quality and lacks stability and accuracy of digital recognition are solved.

The invention discloses a handwritten digit recognition method and device on the basis of image covariance characteristics. The method includes: extracting covariance matrixes of N handwritten digit training images, and using the covariance matrixes as training samples, wherein N is a natural number; utilizing a support vector machine to train the training samples to obtain classifier model coefficients, wherein the kernel function adopted by the support vector machine is a matrix kernel function for calculating similarity among the covariance matrixes; utilizing the classifier model coefficients, the matrix kernel function and the training samples to build a classifier model; utilizing the classifier model to recognize handwritten digit images which need to be recognized. The support vector machine adopting the matrix kernel function can be used for processing the image covariance characteristics, therefore the generated classifier model coefficients are more accurate, and recognition rate of handwritten digits is improved further.

The invention relates to the technical field of image recognition, in particular to a handwritten numeral recognition method based on feature dimension reduction, which comprises the following steps:obtaining a handwritten numeral image for recognition, and obtaining a preprocessed handwritten numeral image after image preprocessing; extracting handwritten numeral features in the preprocessed handwritten numeral image through feature extraction processing; through feature dimension reduction processing, reducing the feature dimension of the handwritten numeral features, and obtaining low-dimension handwritten numeral features; and performing feature recognition processing on the low-dimensional handwritten numeral features to obtain a handwritten numeral recognition result. According to the handwritten numeral recognition method based on feature dimension reduction, feature dimension reduction processing can be carried out on handwritten numeral features so as to reduce the difficultyand complexity of feature recognition, and therefore the recognition effect and efficiency of handwritten numerals can be improved.

The invention discloses a depth learning based smart instrument detection method. The method includes taking a picture of an instrument through a camera, performing preprocessing on the image throughan IPC and reading instrument numerical value through image number recognition technology; performing analysis on the read instrument numerical value data and making result prediction and judgment; sending an instrument change alarm if instrument damage is judged through analysis and taking corresponding halt treatment; sending an alarm and stopping equipment operation immediately if the instrument works normally but the numerical value surpasses a safety range of the instrument; performing smart learning on history data and previous judgment and returning to the system, improving data analysis and system pre-judgment continuously, improving the pre-judgment accuracy through uninterrupted machine self learning. The method provided by the invention has advantages of high smart level and good pre-judgment capability. Besides, the pre-judgment method is improved continuously through history data and history judgment and the smart pre-judgment capability is improved.

The invention discloses an intelligent paper test paper total score identification method and system, which are applied in the field of teacher test paper scoring for improving efficiency , in the method, numbers are accurately segmentedthrough a segmentation algorithm, a unit digit and a ten digit of a score of a corresponding question number in a score region picture are segmented and marked toobtain a marked score region picture, and real numbers are restored. The method is applied to the specific field of paper scoring of teachers and corresponds to a question number score extraction mode. In order to help a teacher group to increase the speed in paper scoring and reduce the burden and pressure of the teacher group, on the premise of changing the score of each question, the corresponding question number score of a score calculation column is recognized through a handwritten numeric recognition neural network, and the scores are added to obtain a total score. A total score calculation result is obtained through rapid identification and rapid calculation, the paper scoring speed is greatly increased, and therefore secondary operation is avoided.

The invention discloses a gas meter character wheel number identification method, relates to a number identification method, and is particularly suitable for carrying out centralized identification ondigital pictures with uniform styles on a gas meter character wheel. The method comprises the following steps: acquiring a digital image from a picture, and identifying the digital image through a trained neural network; When a digital image is acquired, firstly, determining a red area and a black area on a photo, dividing then the photo into single numbers, and acquiring three-dimensional digital images after processing; And training the neural network model by using the three-dimensional digital image, identifying the to-be-identified digital image, and fusing identification results of several neural networks to complete identification. The method comprises the following steps: firstly, identifying a red region on a photo, and quickly positioning a region where a digital image is located; During the period of obtaining the digital image, unrecognized photos can be removed; The to-be-identified digital area adopts secondary positioning, so that the obtained result is more accurate; Different samples are adopted as a training set in different dimensions, three network models with different recognition effects are obtained for recognition, and the recognition accuracy is improved after recognition results of the three networks are fused.

The invention relates to a character wheel image digital recognition method with a fuzzy state, and the method specifically comprises the following steps: obtaining a training set of a state extractor, and obtaining a training set of a classifier; training the state extractor by using the training set of the state extractor, and training the classifier by using the training set of the classifier;and inputting a to-be-identified sample into the trained state extractor to obtain a state variable, inputting the state variable into the trained classifier, and outputting a classification result. According to the invention, effective identification processing is carried out on continuously changing images or images without clear classification boundaries, for example, character wheel image number recognition, the constructed state extractor is used for extracting the characteristic state of a to-be-recognized sample to obtain a state variable, and the to-be-recognized sample is sent into aclassifier corresponding to the membership degree according to the membership degree obtained after the state variable passes through a membership function so that the to-be-recognized sample is correctly classified.

The invention discloses a character recognition method of a convolutional neural network based on RISC-V. The method includes: firstly building a neural network training environment, and then designing a six-layer handwritten number recognition network based on an mnist data set; inputting a picture, performing training to set accuracy, and storing a weight and offset parameter file; setting a custom instruction type and format of the RISC-V instruction set, designing an instruction according to a reserved instruction coding space, and designing a specific hardware acceleration circuit according to the set instruction format and an SIGMOD function implementation method; importing a hardware acceleration circuit, a weight and a bias parameter file into the FPGA, building an RISC-V development environment, and completing the addition of a custom instruction; and inputting the picture, and identifying the numbers in the picture by using a hardware acceleration circuit. According to the invention, instruction-level acceleration of specific operation is realized, partial neural network reasoning operation is accelerated, and handwritten numbers in a picture can be quickly identified.

Methods for discovery of a Markov boundary from data constitute one of the most important recent developments in pattern recognition and applied statistics, primarily because they offer a principled solution to the variable/feature selection problem and give insight about local causal structure. Even though there is always a single Markov boundary of the response variable in faithful distributions, distributions with violations of the intersection property of probability theory may have multiple Markov boundaries. Such distributions are abundant in practical data-analytic applications, and there are several reasons why it is important to discover all Markov boundaries from such data. The present invention is a novel computer implemented generative method (termed TIE*) that can discover all Markov boundaries from a data sample drawn from a distribution. TIE* can be instantiated to discover all and only Markov boundaries independent of data distribution. TIE* has been tested with simulated and re-simulated data and then applied to (a) identify the set of maximally accurate and non-redundant molecular signatures and to (b) discover Markov boundaries in datasets from several application domains including but not limited to: biology, medicine, economics, ecology, digit recognition, text categorization, and computational biology.

The invention discloses a gradient image and similarity weight based digit identification method. The method comprises the following steps: inputting an image needing to be identified; according to prior knowledge, determining a candidate area; performing image enhancement and segmentation on the candidate area; pre-determining whether digits exist in the image; performing area marshalling; carrying out template coupling on an original image on the basis of a gradient image and a similarity weight; and updating a template database. According to the invention, the candidate area is determined according to the prior knowledge so that the algorithm efficiency is improved and the misjudgment possibility is reduced. Through pre-determining whether the digits exist in the image, the average identification efficiency by use of the method can be improved by about 10%. Through respectively performing weighting on two correlation coefficients, a final correlation coefficient is obtained and determination identification is carried out, so that the average identification correct rate can be improved by one to two percentage points.

An automatic recognition method of handwritten test data based on CNN and RNN model comprises the following steps: S1, extracting historical data related to handwritten test digital recognition; obtaining data pictures of various test forms through establishing data interface with test data reporting system; S2, expanding the data: expanding the data picture in step 1; 3, training the CNN + RNN model of handwritten test digit recognition; S4, recognizing handwritten test digits: recognizing the digits in the new handwritten test digit picture according to the CNN + RNN model of the handwrittentest digits obtained in step 3; S5, performing on-site verification: after the on-site personnel check, feeding back whether the test number identified by feedback is consistent with the input number. The invention recognizes the handwritten numerals on the test form through the picture processing and the model training, can directly recognize the data string, including the decimal point in the data string, and can realize the automatic recording of the field operation terminal account and the form.

The invention discloses a combined storage bin for automatic unloading and a bin unmanned vending machine including the storage bin. It includes a cabinet with a hollow structure. An opening A is provided on one side of the cabinet and a locking device is provided on the opening A. Functional cabinet door; at least one storage bin is provided in the cabinet, and the storage bin includes a hollow-structured bin body, and the bottom of the bin body is provided with an opening C for a single cargo to fall out. The counter at the bottom of the cabinet body counts; the cabinet is equipped with a master control circuit board to receive the detection signals of all counters, and the wireless communication module is provided to connect with the remote server for feedback; the cabinet is equipped with The digital identification code that allows the mobile terminal device to connect to the remote server. After the mobile terminal device scans the digital identification code, the cabinet door is opened and the cabinet door is closed to complete the counting and settle the bill through the remote server. The invention has a simple structure, occupies a small area, simplifies the structure, improves the space utilization rate and reduces the manufacturing cost.