904 results about "Single character" patented technology

The invention provides an industrial character identification method based on a convolution neural network. The method comprises the steps of establishing character data sets, carrying out data enhancement and preprocessing on the character data sets and establishing a CNN (Convolution Neural Network) integrated model, wherein the model comprises three different individual classifiers, training is carried out through utilization of the model, the training is finished by two steps, a first step is offline training, an offline training model is obtained, a second step is online training, the offline training model is used for initialization, a special production line character data set is trained, and an online training model is obtained; carrying out preprocessing, character positioning and single character image segmentation on a target image; sending the segmented character images to the trained online training model, and probability values of classifying the single target images into classes by the three classifiers in the CNN integrated model is obtained; final decision is carried out in a voting mode, thereby obtaining a classification result of test data. According to the method, characters on different production lines can be identified rapidly and efficiently.

A handheld device 100 with a graphical user interface for entering handwritten text 102. The handheld device includes word and character input areas 104, 106 within a designated input area 108. Icons 110, 112, 114, 116 and 118 are disposed at the right side of the handwriting user interface 102. A scroll bar 120 may be disposed at the right side of the interface display 102. An entry that begins in the word input area 104 is treated as a handwritten word. A handwritten entry that begins in the character input area 106 is treated as a single character and may be one character in character string. Handwritten character entries are each matched against all potential characters.

The invention discloses an automatic character extraction and recognition system and method for a low-resolution medical bill image. The system comprises an image preprocessing module, a field segmenting module, a single character segmenting module and a character recognizing module. The method comprises the steps of image preprocessing, field area recognizing, character string segmenting and character recognizing and verifying. The automatic character extraction and recognition system and method can be better suitable for automatic character extraction and recognition of the low-resolution medical bill image. The information can be fully utilized by performing layout analysis on a bill. For the image of which the image quality is low and the noise and the image resolution influence are very high, a character string is conveniently segmented into single characters through the semanteme of each field area, and then recognition on the image is converted into recognition on the single characters; for example, an invoice number composed of pure numbers can be recognized through a method special for processing an image only containing numbers, and when the invoice number is recognized, the recognizing range is limited within ten numbers from 0 to 9, and therefore the recognition rate can be greatly increased.

The invention discloses a method for detecting and recognizing sensitive characters in natural scene images, which comprises the following steps of obtaining a training sample, wherein the training sample data comprises a self-built natural scene sensitive text data set and a public text data set; the training sample data comprises a self-built natural scene sensitive text data set and a public text data set; establishing and training the model of direct recognition of sensitive characters based on improved DSSD network; obtaining a test sample image in a natural scene, inputting that test sample image into the sensitive character direct recognition model, detecting and recognizing the sensitive character in the test image, and realizing the sensitive word recognition of a multi-directional and complex changing text region under the natural scene. The method of the invention solves the problem of the stability of the character recognition based on the single character, solves the character sample problem, simplifies the recognition process, and greatly improves the detection and recognition speed and the precision of the sensitive character in the natural scene, and can recognize the sensitive words in multi-directional and complex text regions in natural scenes.

The invention provides a character picture verification code identifying method based on identification feedback. The character picture verification code identifying method is characterized by comprising the following steps of: firstly, converting an original colorful image into a gray level image and carrying out binaryzation treatment to obtain binaryzation image data; then, repairing a binaryzation image, removing image edge burrs and filling a central blank; finishing the connection of broken strokes to obtain a repaired image; removing a background and an interference point or line of the repaired image to obtain a noise-free image; thinning the noise-free image to obtain a thinned image with a single pixel; then, cutting the thinned image to obtain a single-character image which only contains a single character; and finally, carrying out normalization on the single-character image and identifying the character. According to the character picture verification code identifying method based on the identification feedback, a corresponding algorithm is designed to process a character picture identification code including various types of noises and having the phenomena that the character is stuck, inclined, rotated, deformed and the like, so as to finish high-efficiency and high-precision identification. The character picture verification code identifying method based on the identification feedback can be applied to verification code identification in an automatic program.

The invention relates to an automatic license plate positioning and recognition method based on a complex background. Images are acquired in a way that traffic polices or management personnel use digital cameras to take automobile pictures at any time and under various operating conditions instead of acquiring at certain fixed positions or intersections, and automatic license plate positioning and recognition are realized aiming at the random conditions. The method includes the steps: acquiring license plate images; converting color images; acquiring marginal information of the license plate images; acquiring alternate license plate zone images; precisely positioning license plate character zones; subjecting characters to single-character segmentation; and finally, using a BP (back propagation) neural network to realize recognition of license plate characters. The automatic license plate positioning and recognition method based on the complex background has the advantages of high accuracy and the like.

The invention discloses a character detection method and device based on deep learning. The method comprises the steps: designing a multilayer convolution neural network structure, and enabling each character to serve as a class, thereby forming a multi-class classification problem; employing a counter propagation algorithm for the training of a convolution neural network, so as to recognize a single character; minimizing a target function of the network in a supervision manner, and obtaining a character recognition model; finally employing a front-end feature extracting layer for weight initialization, changing the node number of a last full-connection layer into two, enabling a network to become a two-class classification model, and employing character and non-character samples for training the network. Through the above steps, one character detection classifier can complete all operation. During testing, the full-connection layer is converted into a convolution layer. A given input image needs to be scanned through a multi-dimension sliding window, and a character probability graph is obtained. A final character region is obtained through non-maximum-value inhibition.

The invention relates to a character extracting method in a digital video based on character segmentation and color cluster, which comprises the following steps: (1) character segmentation: utilizing the characteristic differences of a character area and a character interval area to carry out vertical projection to segment images in the character area, namely, segmenting each row of area image containing a plurality of characters into a plurality of subarea images only containing a single character so as to reduce the post operating and treating difficulties and improve the identifying accuracy rate of OCR; and (2) character extraction: firstly, using the character color characteristic in the image to cluster colors, finding out an image layer containing maximum character information as a target image layer, and deleting the background area; and then, using the communicating characteristics of the characters to analyze a communicating area of the target image layer, and removing non-character areas to obtain such three results as single character images, an integral image of the character area and an integral image spliced by the single character images respectively, wherein all the three results are input to an OCR system to be identified, and the latter two results use the semantic processing function of the OCR and can accurately determine the characters with similar forms according to the context to improve the identifying effect.

A handwriting recognition method and a handwriting recognition device are provided to recognize a character sequence continuously inputted by a user for convenience. The present method comprises steps of calculating various features of the inputted character sequence which include single character recognition accuracy features and space geometry features of different stroke combinations in the inputted character sequence, calculating segmentation reliabilities of respective stroke combinations in different segmented patterns by using a probabilistic model in which coefficients of the probabilistic model are estimated by a parameter estimation method through sample trainings, recognizing characters in different writing patterns by using a multiple-template matching method when performing single character recognition of the stroke combinations, searching for the best segmentation path and conducting post-processing to optimize the recognition results. The present method and device have advantages of simple structure, low hardware requirement, fast recognition speed and high recognition accuracy and can be implemented in an embedded system.

The invention discloses an estimation method for a dressed body 3D model in single character image. The method comprises the steps of firstly, establishing a 3D model training sample based on a real character, secondarily, studying a varied function between a human body posture, a figure and clothes on the training sample, and calculating an average figure of various real character figures, then selecting a single image containing a dressed body image, enabling a user to select a clothes type and the positions of a plurality of joints, enabling a computer to initialize a three-dimensional posture of the dressed body image according to the positions of the joints and provide the clothes type selected by the user and the average figure for the three-dimensional posture so as to obtain an initial three-dimensional model of the body image, and lastly automatically adjusting the figure and posture of an initialized three-dimensional model according to the outline of the dressed body image to enable the three-dimensional model to conform with the shape of a targeted character. The estimation method for the dressed body 3D model in the single character image only need the single character image to establish the three-dimensional dressed character model and is wide in application range and simple and convenient to operate.

A system and method for providing virtual spaces, where a character associated with a user can be manifested within instances of a plurality of the different virtual spaces. Since a single character can be manifested within instances of different virtual spaces, the character can be transferred by the corresponding user between instances of different virtual spaces and controlled by the user to interact with the different virtual spaces. When the user transfers the character between instances of different virtual spaces (and/or different types of virtual spaces), various aspects of the character may persist between the different virtual spaces (and/or the different types of virtual spaces). This may provide an enhanced continuity to the character between the different virtual spaces.

The invention discloses a video stream-based license plate recognition method, comprising the following steps: S1, detecting triggering, namely defining a rectangular region of interest, obtaining a binary image in the rectangular region of interest by a frame difference, and judging whether a vehicle passes through the region according to the result of the binary image; S2, license plate location, namely carrying out license plate detection in the detection region in the image, and cutting a detected license plate out to obtain a license plate image; S3, license plate pretreatment, namely carrying out license plate inclination correction and accurate license plate cutting on the license plate image; S4, license plate character cutting, namely cutting the character on the license plate image, and decomposing the character into a single character image; S5, license plate character recognition; S6, repeating operations from the steps S2 to S5 on each frame of image to obtain a plurality of license plate recognition results; and S7, result voting. By the video stream-based license plate recognition method, license plate recognition is carried out on the basis of a plurality of frames of images. Thus, the risk caused by a recognition error of a single image is greatly reduced, and the overall recognition rate of the license plate is improved.

The embodiment of the invention provides a method and device for matching regular expressions. The method comprises the following steps of: firstly, inputting a message to be matched and a DFA (Deterministic Finite Automaton) state table, wherein the DFA state table comprises a state transition table including transition relationships between all state addresses and each state in the matching process of the regular expressions; secondly, judging a data type corresponding to the present state, wherein the data type includes a single-character Char type and a multi-character Str type, and the data corresponding to the Str type are a plurality of continuous characters; if the data type is the Str type, matching a plurality of character values in the current state of the message to be matchedand the matching condition, and when matched, shifting to the next state satisfying the matching condition; if the data type is the Char type, matching a single character value at the current state in the message to be matched and the matching condition, and when matched, shifting to the next state satisfying the matching condition; and when the next state is an accept state, finishing the matching process and outputting a success result of matching. The method for matching regular expressions has high matching speed and high efficiency and the space occupied by the DFA items is small.

A method and system for identifying a matching license plate. A pair of full or partial plate numbers to be matched can be received. The pair includes a target plate made up of a first set of characters and a suspect plate made up of a second set of characters. A set of number-order preserved sub-plates can be generated by adding and subtracting a single character from the second set of characters. Utilizing a site specific license plate matching table (P-table), a matching score between each of the sub-plates and the target plate may be calculated as part of a string correlation calculation. A maximum value for each of the sub-plate matching scores can then be calculated. The maximum value can be normalized to a 0˜100 scale and used as a final matching score. When the final matching score exceeds a predetermined threshold, an alert may be issued.

A method for manufacturing a semiconductor device using a photomask and optical lithography is disclosed, wherein circular patterns on the semiconductor wafer are formed by using circular patterns on the photomask, which is manufactured using a charged particle beam writer. In one embodiment, circular patterns of varying sizes have been formed on the photomask using a single character projection (CP) character, by varying the charged particle beam dosage. A method for fracturing circular patterns is also disclosed, either using circular CP characters or using VSB shots wherein the union of the plurality of VSB shots is different than the set of desired patterns.

The invention discloses a license plate recognition method and system, and the method comprises the following steps: collecting a vehicle image; carrying out the filtering of the vehicle image, and carrying out the edge detection of the vehicle image through a Canny operator, so as to convert the vehicle image into an edge binary image; scanning the edge binary image, so as to obtain license plate candidate regions according to the edge spacing and edge jump; locating the license plate in the license plate candidate regions according to the vertical edge features and morphological expansion of the license plate; searching a plurality of characters from the license plate according to the contour features of a single character, and removing the pseudo characters in the characters according to a license plate template; carrying out the recognition of the characters through an SVM classifier, and determining the license plate number of a vehicle according to a recognition result. The method can carry out the precise positioning and character segmentation of the license plate, and improves the character recognition rate of the license plate.

The invention relates to the field of digital image processing and segmenting and the field of tobacco authenticity detection and provides a mobile platform tobacco laser code intelligent identification method and device based on deep learning. Firstly, a laser code image on the surface of a tobacco carton package is obtained through a terminal camera, and binaryzation is conducted on the image by combining the Otsu method and the local thresholding technology; the laser code image is corrected, so that a laser code character zone is located in a rectangular zone parallel to a screen; the laser code image is segmented into single character images through character segmentation; an established deep convolutional neural network is trained by adopting segmented laser code characters, the trained neural network is transplanted to a mobile platform (based on Android) client, and character identification is conducted on a tobacco laser code by adopting the deep convolutional neural network; tobacco authenticity can be judged in time through comparison with a background tobacco coding rule.

The invention provides an entity identification method based on Chinese electronic medical records, and relates to the technical field of medical entity identification. In order to overcome the defects of the lack of a public Chinese electronic medical record annotation corpus in China currently, by constructing and managing a medical dictionary, a semi-automatic corpus annotation method is put forward, and the complexity of manual annotation is reduced. Secondly, the problems are solved that existing electronic medical record entity recognition methods based on characteristics mostly aim at ordinary texts or general electronic medical record texts, and unique characteristics of the Chinese electronic medical records are not considered. By means of the method, besides basic characteristicsof the general text, the unique chapter information characteristics of the Chinese electronic medical records are also extracted; core word characteristics obtained by counting character frequenciesand word frequencies are added into extension characteristics after the collected dictionary is subjected to single-character and word segmentation, a relationship of words is also added to the extension characteristics by clustering word vectors, and the accuracy of the entity identification of the Chinese electronic medical records is effectively improved.

A reduced size and split keyboard arrangement is provided for a portable computing apparatus. The arrangement of the keys, together with mapping of multiple characters to the keys allows reduction of the keyboard to a smaller footprint of a handheld device. Keys may have a single character or multiple characters mapped thereto. For the keys with multiple character mapping, the selection of the character may be by done by simultaneous or near-simultaneous pressing of adjacent physical keys. In one embodiment, disambiguation software may be employed to determine selection of a character from a physical or virtual key with multiple characters assigned thereto.

The embodiment of the invention relates to the communication field, and provides a distributed service-refusal detection defense method and network equipment thereof. The method comprises the following steps: the number N of reached data packages in unit time is counted and compared with a threshold value; when N is not lager than the threshold value, a combination entropy of the reached N data packages in unit time, the mean value of the combination entropy and the fluctuation range of the combination entropy of the reached data packages in unit time are acquired; when N is larger than the preset threshold value, the combination entropy of the reached N data packages in unit time is acquired, and whether the offset of the combination entropy of the reached N data packages in unit time to the mean value of the combination entropy is larger than the fluctuation range of the combination entropy is judged, if so, attack occurrence is determined. The distributed service-refusal DDos attack is detected by using the combination entropy of different parameters of the data packages so as to avoid the shortcoming of high rate of missing report caused by single-character faked by the attackers, thus improving the accuracy in detecting the DDoS attack.

The invention discloses an automatic extracting method for characters on a license plate, which is used for processing a license plate image by a combined global and local binarization method, wherein the global binarization processing is carried out for the license plate image by an Ostu method, the characters and background on the license plate can be accurately separated; and each split single character image is processed by the self-adaptive secondary local binarization method, and the integrity of the character details can be ensured. The method has small operand, high realizability and strong adaptability, can adapt to the license plate images under different environments and illumination conditions, and highlighting the character characteristic can improve the identification accuracy rate of the subsequent characters. The method can be used in an automatic identification system of new vehicle license plates based on digital video/image processing, and for realizing the automatic identification of vehicle license plates in the system.

The invention relates to an input method supporting multimode automatic switching. The method comprises the following steps of: analyzing an input mode of a user according to first stroke information input by the user, and if input is handwriting input, entering a handwriting input mode, tracking and identifying handwriting input in real time and completing input of a single character; if the input is keyboard input, entering a keyboard input mode, feeding back a user key pressing event and recording the input track of the user, performing secondary analysis on subsequent input of the user, and if the subsequent input is handwriting input according to the secondary analysis, clearing the historical marks of the keyboard input and entering the handwriting input mode; and if the subsequent input is keyboard input according to the secondary analysis, keeping the keyboard input mode. The invention further provides an input system supporting multimode automatic switching. Particularly specific to the problem of probable wrong judgment caused by the input of short strokes, all inputs which are determined as key trigger events are further determined and restored, so that wrong judgment is avoided to the maximum extent.

The invention provides a character recognition method based on deep learning. The character recognition method based on deep learning comprises the steps that a deeper multilayer convolution neural network structure is designed, and each character acts as a class; the convolution neural network is trained by using a back propagation algorithm so as to recognize the single character, the target function of the network is minimized in a supervised way and a character recognition model is obtained; and finally the most possible words are found out of the dictionary by adopting a viterbi algorithm according to the existing recognized characters. One input is given, candidate characters are obtained by performing sliding window scanning firstly and then the most possible words are found out of the candidate characters in testing. The character features are learned by utilizing the deeper convolution neural network, the method has robustness in character color, size, illumination and fuzziness, and relatively high accuracy of character recognition and word recognition can be maintained.

State machines used to order and select matching operations for determining whether an input string matches any of at least one regular expression are configured by (1) accepting the set of regular expression(s), and (2) for each of the regular expression(s) of the set accepted, (A) identifying any look-ahead type strings within the given regular expression, (B) identifying any sequential type strings within the given regular expression, (C) partitioning the regular expression based on any identified simple strings, any identified look-ahead type variable strings, and any sequential type variable strings to generate partitioned parts of the given regular expression, (D) reordering the partitioned parts of the given regular expression using optimization policies to generate reordered partitioned parts of the regular expression, and (E) configuring nodes of a state machine corresponding to the given regular expression, by recording configured information of the nodes on a tangible storage medium, using (i) an order of the reordered partitioned parts of the regular expression, and (ii) a string type of the partitioned parts of the regular expression. Once configured, the state machines may accept an input string, and for each of the regular expression(s), check for a match between the input string accepted and the given regular expression using the configured nodes of the state machine corresponding to the given regular expression. Checking for a match between the input string accepted and the given regular expression using configured nodes of a state machine corresponding to the given regular expression by using the configured nodes of the state machine may include (1) checking detection events from a simple string detector, (2) submitting queries to identified modules of a variable string detector, and (3) receiving detection events from the identified modules of the variable string detector.

The invention discloses a license plate positioning identification method based on a support vector machine and a convolutional neural network. The method comprises the following steps of using the weighted grayscale algorithm to carry out graying processing and histogram equalization on an original color license plate image so as to acquire an equalized grayscale image; carrying out edge detection on the equalized grayscale image and acquiring an edge image; carrying out binarization on the edge image; using a morphology to operate the license plate image so as to process and acquiring a license plate candidate area image; using a SVM discrimination model to classify and acquiring a license-plate accurate positioning image; using a vertical and horizontal projection combination method tocarry out image segmentation and acquiring a single character image; using a normalized function to carry out normalization processing on each license plate character; and using a convolutional neuralnetwork model license plate character identification model to identify and outputting and displaying an identification result. In the invention, the robustness of a license plate character identification algorithm is effectively increased, and the method has a high identification rate and is suitable for the license plate positioning and identification under a complex background.

The invention discloses an extremum connected domain based character detection method in a natural scene image. The method comprises the steps of acquiring the natural scene image firstly, and extracting a maximally stable extremal region separated from the natural scene image, wherein the output of the separated maximally stable extremal region is a series of non-overlapped regions, and each region is a connected component; after obtaining the connected components, extracting various features of the connected components, wherein the feature combination can well express the connected components; from the character structure, performing intra-character combination firstly, and then performing inter-character combination, wherein the intra-character combination method is for detecting a single character, and the inter-character combination method is for detecting text lines; and finally, analyzing angular point distribution of the text lines, screening to obtain a text region, and demarcating the text region by a rectangular frame. The method starts with character edge features, and has good pertinence in character detection, thereby having higher initiative and accuracy.