216 results about "Image parsing" patented technology

Intelligent multi-scale image parsing determines the optimal size of each observation by an artificial agent at a given point in time while searching for the anatomical landmark. The artificial agent begins searching image data with a coarse field-of-view and iteratively decreases the field-of-view to locate the anatomical landmark. After searching at a coarse field-of view, the artificial agent increases resolution to a finer field-of-view to analyze context and appearance factors to converge on the anatomical landmark. The artificial agent determines applicable context and appearance factors at each effective scale.

A disclosed facial recognition system (and method) includes face parsing. In one approach, the face parsing is based on hierarchical interlinked multiscale convolutional neural network (HIM) to identify locations and / or footprints of components of a face image, The HIM generates multiple levels of image patches from different resolution images of the face image, where image patches for different levels have different resolutions. Moreover, the HIM integrates the image patches for different levels to generate interlinked image patches for different levels, where interlinked image patches for different levels have different resolutions. Furthermore, the HIM combines the interlinked image patches to identify refined locations and / or footprints of components.

Intelligent image parsing for anatomical landmarks and / or organs detection and / or segmentation is provided. A state space of an artificial agent is specified for discrete portions of a test image. A set of actions is determined, each specifying a possible change in a parametric space with respect to the test image. A reward system is established based on applying each action of the set of actions and based on at least one target state. The artificial agent learns an optimal action-value function approximator specifying the behavior of the artificial agent to maximize a cumulative future reward value of the reward system. The behavior of the artificial agent is a sequence of actions moving the agent towards at least one target state. The learned artificial agent is applied on a test image to automatically parse image content.

A system and method for synthesizing painterly-looking images from input images (e.g., photographs). An input image is first interactively decomposed into a hierarchical representation of its constituent components named parse tree, whose nodes correspond to regions, curves, and objects in the image, with occlusion relations. According to semantic information in the parse tree, a sequence of brush strokes is automatically prepared according a brush dictionary manually built in advance, with their parameters in geometry and appearance appropriately tuned, and blended onto the canvas to generate a painterly-looking image.

The invention provides a comprehensive application method of expressway tunnel driving image acquisition information. video information is acquired by infrared shooting, changes in tunnel driving images and their backgrounds caused by lighting of different lamps can be maximally avoided, difficulty of late image processing can be reduced, feature information of the images is acquired by subjecting the acquired video information to image processing, and finally the feature information is encoded as a control parameter to adjust LED tunnel lamp lighting brightness. The method has the advantages that accuracy is high, tunnel lamps are turned on when vehicles get close and are turned off when the vehicles leave, and energy is effectively saved at the premise of ensuring tunnel driving safety; during detecting the tunnel vehicle information by image analysis, information output through the method necessarily include speeds of vehicles at a monitoring site, quantity of the vehicle, passing time, feature images, vehicle types, license plates and the like, the information is stored, traffic police officers are assisted by data mining in looking into violations of laws, assisting the criminal police in querying the information of suspects, including the time of their vehicles passing a tunnel and the license plates.

Intelligent multi-scale image parsing determines the optimal size of each observation by an artificial agent at a given point in time while searching for the anatomical landmark. The artificial agent begins searching image data with a coarse field-of-view and iteratively decreases the field-of-view to locate the anatomical landmark. After searching at a coarse field-of view, the artificial agent increases resolution to a finer field-of-view to analyze context and appearance factors to converge on the anatomical landmark. The artificial agent determines applicable context and appearance factors at each effective scale.

Provided are a neural network system, and an image parsing method and device based on same. The method comprises: acquiring, by a neural network system, a first visual feature vector x of a target object; according to the first visual feature vector x of the target object, determining, by a shielding estimation sub-network of the neural network system, a shielding mark vector xo of a shielded portion of the target object; according to the first visual feature vector x and the shielding mark vector xo, determining, by a shielding completion sub-network, a third visual feature xc of the target object after complementing the shielded portion; and according to the third visual feature vector xc, determining, by a resolution sub-network, class mark images of all the parts of the target object, and combining the class mark images of all the parts to obtain a parsing result of the target object. The method can solve the image parsing problem when a target object is shielded, and can accurately obtain the parsing result of the target object even if the target object is shielded.

A disclosed facial recognition system (and method) includes face parsing. In one approach, the face parsing is based on hierarchical interlinked multiscale convolutional neural network (HIM) to identify locations and / or footprints of components of a face image. The HIM generates multiple levels of image patches from different resolution images of the face image, where image patches for different levels have different resolutions. Moreover, the HIM integrates the image patches for different levels to generate interlinked image patches for different levels, where interlinked image patches for different levels have different resolutions. Furthermore, the HIM combines the interlinked image patches to identify refined locations and / or footprints of components.

The invention discloses an intelligent robot-oriented man-machine interaction method and an intelligent robot-oriented man-machine interaction device. The intelligent robot-oriented man-machine interaction method comprises an image parsing step, an interaction decision-making step, and an interaction output step. According to the image parsing step, an acquired image sequence related to a target in a preset time is parsed to acquire the behavior information of the target. According to the interaction decision-making step, the active interaction decision-making is executed according to the behavior information of the target, and an active interaction decision-making instruction is generated according to a decision result. According to an interaction output step, actively-interacting multi-mode data is generated and output according to the active interaction instruction. Compared with a conventional man-machine interaction method based on visual sense, the intelligent robot-oriented man-machine interaction method is advantageous in that the accurate prediction of the target behavior is realized, and the effective and individualized interaction between the intelligent robot and the user is realized according to a behavior prediction result, and therefore the user experience of the intelligent robot and user stickiness are improved.

The invention relates to a data transmission method, device and system, computer equipment and a storage medium. The reading server analyzes the unread image data in the FTP server database to obtainimage analysis data, and then sends the image analysis data to a message system for storage, so as to subscribe a consumption group of a topic where the image analysis data is located to obtain the image analysis data from the message system. The image data collected by different types of image collection devices can be stored in the FTP server database; image data acquired by different image acquisition devices can be stored and analyzed through the method and then sent to a message system. Therefore, when the image data needs to be acquired, different access adapts do not need to be customized for different types of image acquisition equipment, and the image data of the corresponding equipment can be acquired only by subscribing from the message system, so that the image transmission process is greatly simplified, the SDK development cost is reduced, and the flexibility of the image data transmission process is improved.

The invention provides a house type vectorized data obtaining method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining a house type image; analyzing the scale proportion of the house type image, wherein the scale proportion represents the proportion value of the house type size of the actual house and the house type size in the house type image; respectively extracting wall contour data and bay window contour data in the house type image; and determining vectorization data according to the scale proportion, the wall contour data and the bay window contour data. In the implementation process, the scale proportion of the house type image is analyzed through an electronic equipment executive program; wall contour data and bay window contour data are respectively extracted in the house type image; the vectorized data corresponding to the house type image is determined, and the vectorized data corresponding to the house type image is extracted by using the electronic equipment executive program, so that the error rate of manual extraction is reduced, and the efficiency of extracting the vectorized data from the house type image is effectively improved.

In an image processing apparatus, an image acquiring section acquires a group of candidate images for use in creating a recommended composite image to be proposed to a user; an image analysis information acquiring section acquires image analysis information on each image of the group of candidate images; a theme determining section determines a theme of the group of candidate images based on the image analysis information on each image of the group of candidate images; a template selecting section selects a template for use in creating a recommended composite image based on the theme of the group of candidate images; and a layout section creates a recommended composite image in which images of the group of candidate images are laid out in the template.

The invention provides a video signal black-screen-free switching processing method. The method comprises the following steps: receiving a video signal; performing time sequence monitoring on the received signals according to the current application configuration and the EDID information of the rear-end display device; when it is monitored that the input time sequence is changed, the output picture is locked, and then the effect of locking the last frame of picture when the signal is lost or switched is achieved; when it is monitored that the input time sequence is stable, image resolution conversion is conducted according to the configured resolution; and displaying the image after resolution conversion. According to the invention, seamless switching and advanced image resolution conversion technologies can be combined, switching actions without black screen, screen turn-off and flicker can be provided during video signal switching, and the requirements of high-quality and high-resolution video demonstration are met.

The invention relates to a method and system for regularly detecting an X-ray machine image resolution testing body, and belongs to the technical field of image resolution. The method includes the following steps of S1, image testing, S2, zone locating, S3, TEST1 detecting algorithm testing, S4, TEST2 detecting algorithm testing, S5, TEST3 detecting algorithm testing, and S6, TEST4 detecting algorithm testing. The method and system for regularly detecting the X-ray machine image resolution testing body intends to integrate leading edge theories of compute vision and mode recognition and develops an automatic regular checking technology of an X-ray security checking machine based on intelligent image analyzing, and through automatic analyzing of imaging of a standard testing box, whether the performances of the line resolving power, penetrating resolving power, space resolving power, penetrating power and the like of the X-ray security checking machine are up to standard or not are accurately and objectively diagnosed in real time to give expert-level opinions on whether security devices can be normally used or not.

A system and method providing copies of content is disclosed. The system includes a content distribution hub and a remote content distribution device. The content distribution hub receives some content and creates a recordable medium image of the content. The content distribution hub parses the recordable medium image of the content into data blocks and broadcasts the data blocks to the remote content distribution device. The remote content distribution device receives the data blocks and generates a second recordable medium image of the content using the received data blocks. The content distribution device is adapted to record a copy of the second recordable medium image of the content into a unit of recordable media.

The invention provides an edge computing gateway based on visual identification. The edge computing gateway comprises an image acquisition module, an image analysis module, a data conversion module, amicroprocessor, a protocol analysis module and a data transmission module. The image acquisition module comprises a panel image acquisition camera group and an environment image acquisition camera. The image analysis module analyzes to obtain parameter information and alarm information of each device. The microprocessor is used for processing and analyzing the converted parameter information andalarm information to obtain an abnormal result and an operation action of the corresponding equipment. The data transmission module is used for sending the abnormal result converted into the preset protocol and the equipment operation action to the corresponding equipment. According to the scheme, the trouble of a conventional edge computing gateway on a closed protocol is made up, equipment whichis isolated from each other due to the fact that a communication protocol is not opened or no communication interface exists before is communicated, monitoring of the whole workshop can be completed,and the workload of staff inspection is greatly reduced.

The invention relates to the field of intelligent cabins, and provides a code scanning method based on a vehicle-mounted camera, and the method comprises the steps: collecting a first image through the vehicle-mounted camera; analyzing a target coding pattern contained in the first image; and when the analysis information is not obtained, based on the image data obtained by the vehicle-mounted camera, collecting other images adjacent to the first image in time relation for analysis so as to try to obtain the analysis information in the target coding pattern. The code scanning method based on the vehicle-mounted camera can solve the problems that operation is complex, consumed time is long, the scanning success rate in a vehicle moving scene is low, attention of a driver can be distracted easily through manual operation of a terminal, and traffic jams are caused easily when intelligent terminal equipment is used for code scanning in a driving scene. The efficiency of the code scanning operation in the driving scene and the driving safety are improved.

The invention discloses a tumor metastasis unit counting method based on quantum dot spectrum analysis and image parsing, which can be used for rapidly realizing the in-situ imaging of a plurality of components in a tumor microenvironment, the quantitative calculation of each component and the quantitative counting of tumor metastasis units. According to the method, the shortcomings of conventional and modern methods are overcome; the method is high in detection speed and stability; manual calculation is eliminated, so that individual judgment errors are avoided; a metastasis unit consisting of the components can be effectively, conveniently, rapidly and quantitatively calculated, and the development of individualized tumor treatment is favorably promoted.

A controller (print control unit) comprises an image analyzing part automatically analyzing an image for printed matter and an inference part adjusting the value of a control parameter related to a printing Machine with fuzzy inference. The inference part acquires the feature of the image such as "solid", "halftone" or "light" with the result of analysis by the image analyzing part, and decides each control parameter in response to the feature of the image. The printing machine makes printing output with the adjusted control parameter transmitted from the controller. The feature of the image may alternatively be supplied as an instruction from an operator through a feature specifying part. Thus, a print control unit capable of excellently and simply adjusting the control parameter for the printing machine in response to the feature of the printed image is provided.

The invention provides a glass-plastic mixed lens comprising a first lens, a second lens, an aperture diaphragm, a third lens, a fourth lends, a 5th lens, a 6th lens and an optical filter arranged insequence from the object side to the image side; the 1st to 4th lenses respectively refer to a spherical surface glass lens; the 5th and 6th lenses are non-spherical surface plastic lenses; the firstlens refers to a meniscus lens with a negative focal power and a convex object surface; the second lens is a meniscus lens with a positive focal power and a convex object surface; the third lens is adual-convex lens with a positive focal power; the 4th lens is a dual-concave lens with a negative focal power; the 5th lens is a lens with a positive focal power and a convex image surface; the 6th lens is a dual-concave lens with a negative focal power. The glass-plastic mixed lens employs two non-spherical surface plastic lenses to improve the imaging sharpness, thus reducing the aberration influences, obtaining better image parsing ability, and satisfying unmanned plane outdoor aerial photography needs.

The invention is suitable for the technical field of data processing, and provides an abnormal object identification method and terminal equipment, and the method comprises the steps: receiving the object data of a to-be-recognized target object; respectively extracting an image feature value of each object image, and matching the image feature value with a historical feature value of each historical image in a database to obtain a matching result between the image feature value of the object image and the database; if the matching result is that the matching is successful, respectively importing each object image of the target object into a preset image analysis model to obtain object attribute information of each object image; and outputting an exception identification result of the target object based on all the object attribute information. According to the invention, online authentication is realized, the identification efficiency is improved, the identification difficulty is reduced, and the labor cost of the identification process is further reduced.

The invention provides a device and a method for browsing images. The device comprises: an image capturing unit for capturing images; an image parsing unit for parsing the images captured by the image capturing unit to obtain metadata of the images; a database for storing the metadata of each image parsed by the image parsing unit; a list obtaining unit for generating a sequence list of the images to be browsed from the metadata of each image stored in the database based on a browse condition; and a display unit for displaying the browsed images based on the sequence list generated by the list obtaining unit.

Techniques for spatial semantic attribute matching on image regions for location identification based on a reference dataset are provided. In one aspect, a method for matching images from heterogeneous sources is provided. The method includes the steps of: (a) parsing the images into different semantic labeled regions; (b) creating a list of potential matches by matching the images based on two or more of the images having same semantic labeled regions; and (c) pruning the list of potential matches created in step (b) by taking into consideration spatial arrangements of the semantic labeled regions in the images.

The invention discloses a coal gangue separation method based on deep vision, and relates to the technical field of robot vision. The method comprises the steps of taking ore color images of an ore separation production line as input, utilizing an ore segmentation convolutional neural network for conducting segmentation to obtain an ore area segmentation result, conducting ore area cutting on theore color images according to the ore segmentation result to obtain a plurality of ore area slices, then utilizing an ore classification convolutional neural network to output coal briquettes or coalgangue, conducting image analysis on the ore area slices to determine ore information, and finally automatically separating the coal briquettes and the coal gangue. By means of the method, rapid detection and precise recognition can be achieved, by estimating the volume class and relative position of ore, accurate guide information is provided for automatic separation of the coal gangue, and the separation efficiency is high. The two convolutional neural networks are customized special for an ore separation scene, and the coal gangue separation method has the advantages of being small in mass,high in real-time performance, high in detection precision and the like.

The invention relates to the technical field of network equipment installation, in particular to a method for installing network equipment in batch. The method adopts an installation rack, a background system, a switch and a control terminal. The installation rack is used for installing the network equipment; the installation rack is provided with ports for being connected with the network equipment; the ports are connected with the switch by network cables; an image acquisition terminal is also arranged on the installation rack; the image acquisition terminal can acquire images of tags of the network equipment and can transmit the images to the background system by the switch, and the background system can carry out implementation on the tag images by an image parsing module so as to carry out modification on login passwords of the network equipment. According to the method for installing the network equipment in batch, which is provided by the invention, the network equipment can be configured in batch in the deployment, and the problems of high modification workload, complex process and big potential safety hazard when an enterprise deploys the network equipment are solved.

A cell image analysis method and a cell image analysis device are provided. A shape of a cell is favorably observed in a non-invasive manner. One aspect of the cell analysis device according to the present invention includes: holographic microscopy 10, an image generation unit 23,24 configured to generate a phase image of a cell based on hologram data acquired by observing the cell with the holographic microscopy, a first machine learning model storage unit 251 configured to store a cell area machine learning model generated by performing machine learning using training data in which a phase image of a cell generated based on the hologram data is used as an input image and a pseudo cell area image based on a stained image acquired by staining a cytoskeleton corresponding to the phase image is regarded as a ground truth image, and a cell area estimation unit 25 configured to output a cell area estimation image indicating a cell area using a machine learning model stored in the first machine learning model storage unit, wherein the phase image generated by the image generation unit for the analysis target cell is used as an input image.

The invention provides a Unity-based virtual scene shelf display method and system. The method comprises the following steps: establishing a shelf display 3D model set; wherein the shelf display 3D model set comprises a commodity 3D model set and a display carrier 3D model set; acquiring shot image data; identifying and analyzing the image data through a virtual scene shelf display deep learning algorithm to obtain image analysis data; and matching the commodity 3D model and the display carrier 3D model which are consistent with the image analysis data in the shelf display 3D model set, and placing the matched commodity 3D model in the matched display carrier 3D model to complete 3D virtual display reproduction. According to the method, the situation of the display scene is restored in the three-dimensional virtual scene, the restored result pictures and the terminal display videos under various postures can be provided, and the terminal display situation can be comprehensively and accurately controlled for fast-moving consumer enterprises, so that the actual situation of the sales market can be timely and effectively controlled, and a correct decision can be made.

The invention discloses a hyperspectral image water area automatic extraction method, and belongs to the technical field of remote sensing image processing. The method comprises the following steps: firstly, reading a hyperspectral image, analyzing waveband information of data, and calculating a water body index value of each pixel; then, an improved OTSU method is used for carrying out water areaautomatic segmentation threshold calculation, and a water area primary extraction result is obtained; then, on the basis of a water area primary extraction result, extracting a spectrum of water, carrying out spectrum consistency analysis, and obtaining an Euclidean distance between each pixel and a water body spectrum; taking the Euclidean distance as a weighted value, carrying out covariance matrix calculation, and carrying out water area extraction of a constraint energy minimization method on the basis; obtaining a water area secondary extraction result by utilizing an improved OTSU method; and finally, comprehensively analyzing the water area primary extraction result and the water area secondary extraction result to obtain a final water area extraction result. The method can be carried out in a full-automatic mode, priori knowledge and any parameter setting are not needed, and high precision is achieved.

An embedded image processing identification system and method comprises an image sensor, an optical sensor, and a signal processing and control module; wherein the image sensor is used for capturing an image of an external environment; the optical sensor is used for sensing an optical characteristic value about the external environment, in order to generate a sensing signal; and the signal processing and control module comprises a plurality of image parsing calculating methods corresponding to different environment states, and is used for selecting the proper image parsing calculating method to parse the image according to the sensing signal, in order to obtain an identification image about a characteristic object. The embedded image processing identification system utilizes the optical sensor to sense brightness of an ambient light source, so that the calculating methods can be switched smoothly when intensity of the ambient light source changes, so as to adapt to different weathers.