345 results about "Segmentation system" patented technology

The invention discloses a fundus image retinal vessel segmentation method and a fundus image retinal vessel segmentation system based on deep learning. The fundus image retinal vessel segmentation method comprises the steps of performing data amplification on a training set, enhancing an image, training a convolutional neural network by using the training set, segmenting the image by using a convolutional neural network segmentation model to obtain a segmentation result, training a random forest classifier by using features of the convolutional neural network, extracting a last layer of convolutional layer output from the convolutional neural network, using the convolutional layer output as input of the random forest classifier for pixel classification to obtain another segmentation result, and fusing the two segmentation results to obtain a final segmentation image. Compared with the traditional vessel segmentation method, the fundus image retinal vessel segmentation method uses the deep convolutional neural network for feature extraction, the extracted features are more sufficient, and the segmentation precision and efficiency are higher.

The invention provides an unsupervised domain-adaptive brain tumor semantic segmentation method based on deep adversarial learning. The method comprises the steps of deep coding-decoding full-convolution network segmentation system model setup, domain discriminator network model setup, segmentation system pre-training and parameter optimization, adversarial training and target domain feature extractor parameter optimization and target domain MRI brain tumor automatic semantic segmentation. According to the method, high-level semantic features and low-level detailed features are utilized to jointly predict pixel tags by the adoption of a deep coding-decoding full-convolution network modeling segmentation system, a domain discriminator network is adopted to guide a segmentation model to learn domain-invariable features and a strong generalization segmentation function through adversarial learning, a data distribution difference between a source domain and a target domain is minimized indirectly, and a learned segmentation system has the same segmentation precision in the target domain as in the source domain. Therefore, the cross-domain generalization performance of the MRI brain tumor full-automatic semantic segmentation method is improved, and unsupervised cross-domain adaptive MRI brain tumor precise segmentation is realized.

The customer segmentation software according to the present invention automatically finds or creates profiles of prototypical customers in a large e-commerce database. The software matches all existing customer data in the database to one or more of the prototypical customers. The resulting customer segmentation is an effective summarization of the database and is useful for a range of business applications. Applications of the customer segmentation system include the development of customized web sites, the creation of targeted promotional offers and the prediction of consumer behavior.

Remote segmentation applied to a segmentation data mart allows a marketer to create a personalized email campaign for a selected segment of customers. Segmentation data is collected from a plurality of third party sources, imported and cleansed. The marketer may query a data mart with a user-defined rule created with parameters selected from fields available in the data mart. The marketer submits the query and is presented with a count with which the marketer may determine if the segment will be cost effective for the marketing campaign. If the count is acceptable, the query is saved. Later, when the marketer creates the email message for a particular campaign, s / he assigns the segment to the campaign. When the campaign is released, the query extracts email addresses currently meeting the criteria of the query and uses the addresses for distributing the email.

Systems and methods for automated segmentation of anatomical structures, such as the human heart. The systems and methods employ convolutional neural networks (CNNs) to autonomously segment various parts of an anatomical structure represented by image data, such as 3D MRI data. The convolutional neural network utilizes two paths, a contracting path which includes convolution / pooling layers, and an expanding path which includes upsampling / convolution layers. The loss function used to validate the CNN model may specifically account for missing data, which allows for use of a larger training set. The CNN model may utilize multi-dimensional kernels (e.g., 2D, 3D, 4D, 6D), and may include various channels which encode spatial data, time data, flow data, etc. The systems and methods of the present disclosure also utilize CNNs to provide automated detection and display of landmarks in images of anatomical structures.

Described is a bio-inspired vision system for attention and object segmentation capable of computing attention for a natural scene, attending to regions in a scene in their rank of saliency, and extracting the boundary of an attended proto-object based on feature contours to segment the attended object. The attention module can work in both a bottom-up and a top-down mode, the latter allowing for directed searches for specific targets. The region growing module allows for object segmentation that has been shown to work under a variety of natural scenes that would be problematic for traditional object segmentation algorithms. The system can perform at multiple scales of object extraction and possesses a high degree of automation. Lastly, the system can be used by itself for stand-alone searching for salient objects in a scene, or as the front-end of an object recognition and online labeling system.

In a human feature recognition system that is intended to provide substantially real-time recognition of body segments, various methods and structures are provided to facilitate real-time recognition with reduced computation requirements, including a face detection module employing an active boosting procedure and a lazy boosting procedure on a hybrid cascade structure, a human body segmentation module and a boundary matting module. The hybrid cascade structure is in the form of a tree where one type of node represents a strong classifier learned from active boosting, another type of classifier is obtained by low-computation-load lazy boosting, and weak classifiers are obtained from the previous layers.

A campaign is received, at a micro-segmentation system, from an offer provider. The micro-segmentation system is a third-party system. The campaign indicates a set of target attributes and one or more offers corresponding to the set of target attributes. A set of user attributes pertaining to each of the plurality of users is received at the micro-segmentation system. The set of attributes is defined by an attribute knowledge structure. Permission is received at the micro-segmentation system from each of the plurality of users to receive an offer from the micro-segmentation system. Data associated with the micro-segment classification is provided from the micro-segmentation system without user information to the offer provider.

The invention provides a text-subject-model-based data processing method for commodity classification. The method comprises the following steps of: importing Chinese and English vocabulary related to a service into a universal word library of a word segmentation system, and importing white name English words related to the service for brands and common commodity English; further expanding a stop word library of the word segmentation system; segmenting words for a description text part of a commodity, so that each commodity can have a bag of words which is not related to sequence; counting word segmentation results to acquire uncommon vocabulary with high frequency, and thus constructing a preferential word library; and appointing a general classification quantity, setting related parameters, executing quick Gibbs sampling, acquiring potential semantic association, comparing the latent semantic association with the preferential word library, the universal word library and the stop word library respectively, calculating comparison results to obtain the most possible classification of the commodity, and marking the classification by using the bags of words. In consideration of latent semantics, the influence of subjective factors of editorial staff is reduced, so that the commodity classification is accurate.

The invention discloses a segmentation index segmentation method. Including the following steps: read the character stream; identification described the character stream to identify Chinese characters and English characters, as well as an identification number or character; already identified Chinese and English characters or Digital and pre-built 1.1 tree comparison, the sub-set match words; English characters or figures generic fuzzy matching ASCII codes to determine English string or string of digital-term matching the above mentioned English words and string or digital string of words and non-recognition of characters referred to the character stream by order of ranking; The words and figures mentioned in the English string or strings of the sort described in the order of the character stream. The invention also openly segmentation Index segmentation system. The invention provides a cut-word indexing method and system can simultaneously address the precise words, a certain amount of redundant words and word-term problems, enhance the user experience.

The invention discloses a Chinese word segmentation method, which comprises the following steps of: performing word segmentation on a Chinese text according to word semantics, segmenting ambiguous fields and outputting a first text string taking words as units; and identifying and combining Chinese names in the first text string to generate a second text string taking words as units. The ambiguous fields are segmented by combining a dictionary rule method with a statistical method; and the ambiguous fields are segmented and the names are identified by word standard a maximum entropy model in the statistical method. The invention also discloses a Chinese word segmentation system, which comprises a word segmentation module, a name identification module and the like. The method and the system improve word segmentation efficiency and accuracy.

Systems and methods for training semantic segmentation. Embodiments of the present invention include predicting semantic labeling of each pixel in each of at least one training image using a semantic segmentation model. Further included is predicting semantic boundaries at boundary pixels of objects in the at least one training image using a semantic boundary model concurrently with predicting the semantic labeling. Also included is propagating sparse labels to every pixel in the at least one training image using the predicted semantic boundaries. Additionally, the embodiments include optimizing a loss function according the predicted semantic labeling and the propagated sparse labels to concurrently train the semantic segmentation model and the semantic boundary model to accurately and efficiently generate a learned semantic segmentation model from sparsely annotated training images.

The disclosure is directed to techniques for automatic segmentation of a region-of-interest (ROI) video object from a video sequence. ROI object segmentation enables selected ROI or “foreground” objects of a video sequence that may be of interest to a viewer to be extracted from non-ROI or “background” areas of the video sequence. Examples of a ROI object are a human face or a head and shoulder area of a human body. The disclosed techniques include a hybrid technique that combines ROI feature detection, region segmentation, and background subtraction. In this way, the disclosed techniques may provide accurate foreground object generation and low-complexity extraction of the foreground object from the video sequence. A ROI object segmentation system may implement the techniques described herein. In addition, ROI object segmentation may be useful in a wide range of multimedia applications that utilize video sequences, such as video telephony applications and video surveillance applications.

The invention discloses an Internet forum-oriented opinion leader mining method. An opinion leader mining system is involved in the method and comprises a computing center and a database server which communicates with the computing center. The method comprises the following steps of: capturing forum data by using a crawler, and improving data processing real-time property by using message-oriented middleware (MOM); extracting web page information, performing word segmentation by using a Chinese word segmentation system, and filtering spam comments by a spectral clustering method; analyzing text tendency by using an emotional corpus; setting a selection standard value of an opinion leader, and determining the opinion leader; and visualizing a result. By the method, the opinion leader in a forum can be accurately mined, and technical support is provided for related Internet public opinion supervision departments to timely find hot issues and guide the healthy development of Internet public opinions.

The invention discloses a text keyword extraction method based on combination of Word2Vec and word co-occurrence. According to the method, an ICTCLAS word segmentation system is adopted to perform word segmentation and part-of-speech tagging on a text to obtain a vocabulary set; then, the vocabulary set is preprocessed, unreasonable vocabulary combinations are filtered out, and a preliminary candidate set is obtained; the preliminary candidate set is placed in a trained Word2Vec model to obtain a word vector table, the distance between word vectors in the word vector table is calculated, kmeans clustering is performed on the preliminary candidate set to obtain a secondary candidate set of keywords, and a word co-occurrence rate of the secondary candidate set in the preliminary candidate set is obtained according to the word vector distance; and different weight values are given to different vocabulary lengths, corresponding weights are obtained according to the word co-occurrence rateand the vocabulary lengths, ordering is performed according to the weights, and the first m keywords are final keywords. Through the method, the word vectors generated through Word2Vec are adopted toperform clustering, then the text keywords are extracted in combination with word co-occurrence and other basic characteristics, therefore, the extracted keywords are more accurate, and the method canadapt to extraction of keywords of different texts.

The invention discloses a panoramic segmentation method based on a graph neural network. The panoramic segmentation method comprises the following steps: extracting a plurality of target features froma picture; segmenting the head network through an example to obtain a foreground category probability, a background category probability and a mask result of the picture, and semantically segmentingthe head network to obtain a preliminary semantic segmentation result of the picture; processing the new foreground image through the foreground category probability to generate an instance classification result, and extracting a target instance segmentation mask from the instance classification result according to a mask result; processing the new background image through the background categoryprobability and the preliminary semantic segmentation result to generate a target semantic segmentation result; and fusing the target instance segmentation mask and the target semantic segmentation result by adopting a heuristic algorithm to generate a panoramic segmentation result. The invention further discloses a panoramic segmentation system based on the graph neural network, computer equipment and a computer readable storage medium. By adopting the method and the device, the panoramic segmentation effect of the picture can be optimized by utilizing the mutual relation between the objects.

A video segmentation system generates an S-distance measurement that is a representation of the similarity between adjacent frames of a video sequence. The video segmentation system employs frequency decomposition of a direct current (DC) luminance signal of a compressed video sequence. High and low frequency component signatures are generated from a frequency-decomposed signal using wavelet transformation. A cut detector identifies cut transitions from the low frequency component signature. A fade detector identifies fade transitions the high frequency component signature. A dissolve transition detector employs a double frame differencing algorithm to identify dissolve transitions. A video retrieval system likewise generates an S-distance between a query image and a database image. The video retrieval system employs the low and high frequency component signature to generate the S-distance measurement of the similarity between the query image and the database image. The results of the S-distance measurement allow browsing and searching of the similar database images.

The invention relates to an audio editing system. The audio editing system comprises a plurality of initial segmentation devices, a multi-sound track fusion device, an audio clustering device and a re-segmentation device, wherein the plurality of the initial segmentation devices are respectively used for initially segmenting audio streams from a plurality of sound tracks into a plurality of different paragraphs; the multi-sound track fusion device is used for integrating segmentation points of the plurality of the initial segmentation devices, selecting the audio stream of the optimal sound track between every two adjacent segmentation points, further getting a plurality of initially segmented fragments and fusing the plurality of the obtained initially segmented fragments into an uniform audio data file; the audio clustering device is used for performing clustering on the plurality of the initially segmented fragments under supervision based on a hierarchical clustering algorithm and clustering the initially segmented fragments belonging to the same nature to a category; and the re-segmentation device is used for training according to the clustering result of the audio clustering device to get a hidden Markov model corresponding to each type and performing Viterbi alignment segmentation on the uniform audio file to get the audio stream after re-segmentation. The accuracy in final speaker clustering can be improved through a high-precision speaker segmentation system.

The invention provides a blood vessel segmentation method and a system of fundus image. The blood vessel segmentation method of fundus image comprises the following steps: step S1, constructing a neural network training set of fundus image for neural network training; step S2, amplifying the neural network training set to obtain the amplify neural network training set; step S3, constructing a neural network model for blood vessel segmentation of fundus oculi image based on depth learning; step S4, training the neural network model by using the expanded neural network training set; step S5, obtaining a target fundus image, and performing blood vessel segmentation on the target fundus image by using the trained neural network model. The blood vessel segmentation method and the blood vessel segmentation system of the fundus oculi image of the invention can overcome the limitation that the number of the blood vessel segmentation public data sets is low and can be adapted to the fundus oculi image of poor quality. The invention greatly simplifies the network structure, and can effectively improve the segmentation speed without losing the larger accuracy rate.

A Chinese word segmentation method based on navigation information retrieval is characterized in that a word segmentation system is obtained through the steps that a dictionary is loaded, and text code conversion is carried out; segmentation processing is carried out, and a source character string is segmented into a plurality of slightly simpler short sentences; atomic word segmentation is carried out to obtain the smallest morpheme units which cannot be segmented in the short sentences; word forming full-match is achieved with a word-by-word traversal matching method; the matching results are screened to generate a plurality of best results; human names, place names and proper nouns are processed; the dictionary is corrected, and mainly, unlisted new words are added, and properties of the existing words are improved; the processing results of all the short sentences are finally combined to be output. The Chinese word segmentation method has the advantages that content input by a user can be formed into words through the Chinese word segmentation technology, the speed can be optimized, wrongly written characters can be corrected with the words as the basis, and a more suitable result can be provided. With the Chinese word segmentation technology, semantics can be understood by an information retrieval engine better, and the provided result set can be fully adjusted.

A Hidden Markov model is used to segment a data sequence. To reduce the potential for error that may result from the Markov assumption, the Viterbi dynamic programming algorithm is modified to apply a multiplicative factor if a particular set of states is re-entered. As a result, structural domain knowledge is incorporated into the algorithm by expanding the state space in the dynamic programming recurrence. In a specific example of segmenting resumes, the factor is used to reward or penalize (even require or prohibit) a segmentation of the resume that results in the re-entry into a section such as Experience or Contact Information. The method may be used to impose global constraints in the processing of an input sequence or to impose constraints to local sub-sequences.

The invention provides a scenarized semantic comprehension and dialogue generation method and a scenarized semantic comprehension and dialogue generation system. The method comprises the steps of: establishing a user scene model, selecting and determining the field of a current conversation according to a result of a word segmentation system; understanding an interactive content of the round based on a corresponding scenarized user model by use of a selected scene semantic parser; and calling a dialogue generator in a corresponding scene, carrying out a dialogue synthesis and generating a dialogue result after interaction of the round in combination with an intermediate state of dialogue management. According to the scenarized semantic comprehension and dialogue generation method and the scenarized semantic comprehension and dialogue generation system, natural language understanding for a sentence and a phrase is simply and efficiently realized, and the situation that a computer automatically and completely understands a short text is realized. Automatic deep understanding and dialogue interaction of the computer for the sentence or the phrase of a natural language are realized and the purpose that a user needs a machine to understand the interaction language use automatically and accurately is achieved.

The invention relates to a comb tooth type ricinus communis combine harvester with fruit picking and straw fixed length fixed quantity and bundling functions. The combine harvester is reasonable in structure and integrates picking, separating and cleaning of castor seeds with fixed length segmentation, fixed quantity compaction and automatic bundling of straw. A comb tooth picking system and a feed-in disc type header system are located at the front end of a harvester body, a separating and cleaning system and a hob type fixed length segmentation system are located at the middle end of the harvester body, and fixed quantity compaction and bundling systems are located at the rear end of the harvester body; a comb tooth picking device which rotates anticlockwise is adopted for the comb tooth picking system to pick off the castor seeds and send the castor seeds to the separating and cleaning system, the castor seeds after being collected and separated through a grain collection bin are arranged below an outlet of the separating and cleaning system, a divider is arranged in the front of the harvester body, ricinus communis straw is fed into the disc type header system, the cutoff straw is fed into the hob type fixed length segmentation system through a stalk lifting conveyer belt, and an outlet of the hob type fixed length segmentation system is connected with the fixed quantity compaction and bundling systems; all the systems are driven by a power device through corresponding transmission devices.