63 results about "Olleya aquimaris" patented technology

The invention provides a multimode recurrent neural network picture description method based on FCN feature extraction. A multimode model composed of three parts, namely, a recurrent neural network (RNN), a fully convolutional neural network (FCN) and a multimode layer is obtained by training massive images labeled with text description, and automatic generation of text description of any input test image is achieved. By means of the method, image features can be effectively extracted, more detail information of the images can be retained, and the relation between words in the text description and the images can be better established. The method has significant advantages on semantics-based description between image salient targets or scenes.

The invention discloses an urban road network path planning method for the non-global information and improves efficiency of the urban transportation network through balancing the traffic network flow. A path planning method in the prior art can only calculate the best path for a single vehicle or requires more information and assumptions. In response to the above problems, through periodically monitoring state changes of road segments, intelligently learning and evaluating current road segment congestion indexes and constructing a PST model, the A*R algorithm is utilized for path selection, so the road network state is made to be in a flow balancing state. The method is advantaged in that a large number of experiments show that the method can more effectively balance the traffic network flow and alleviate traffic congestion in the scene with relatively a large number of vehicles and has a good performance in reducing the average driving time and the average driving distance, in addition, the method can significantly improve traffic congestion even when relatively lots of vehicles in the road network do not accept path planning navigation.

The invention discloses a panoramic synthesis method for motion video. Wherein, the motion video contains motion foreground and background; every video frame contains motion foreground and part background. The method comprises: using said a plurality of video frames to form a background panorama; extracting the motion foreground to overlay on background panorama. This invention has advantages as it is efficient to sharp motion of foreground or camera, and can show well the process and details of motion object action.

The invention discloses a high-selectivity method for preparing linear-chain aldehyde. According to the method disclosed by the invention, a phosphine-containing organic polymer self-loaded type high-dispersity metal catalyst is used, and olefin hydroformylation reaction is used for preparing the linear-chain aldehyde in a high-selectivity manner. The organic polymer self-loaded type catalyst takes one, two and three of metal Rh, Co, Pd, Ir and Rt as active components and a phosphine-containing organic polymer is used as a carrier, wherein the phosphine-containing organic polymer carrier is formed by co-polymerizing or self-polymerizing a multi-dentate organic phosphine ligand and a single-dentate organic phosphine ligand. The phosphine-containing organic polymer self-loaded type high-dispersity metal catalyst provided by the invention is applied to reactors including fixed beds, slurry beds, bubbling beds, trickle beds and the like. The method disclosed by the invention has very goodpresentation in the olefin hydroformylation reaction and can be used for producing a target product linear-chain aldehyde in the high-selectivity manner.

The invention discloses a sonar target detection method based on the Faster R-CNN. According to the method, characteristic extraction is performed on targets in different complex underwater environments through utilizing the deep learning technology, the method is the major innovation and attempt compared with present sonar target detection methods, limitations of traditional methods are broken, deep characteristics of sonar images at the low SNR can be extracted, and good target detection and identification on the linear targets can be carried out. The sonar target detection and identification network based on the regional accelerated convolutional neural network (Faster-RCNN) is established to carry out target detection for sonar data. The method is advantaged in that good performance oflinear target detection of the sonar images is achieved, feasibility of the deep learning method in sonar target detection is verified, and a new research method is provided for characteristic extraction of the complex underwater acoustic environment data.

The invention discloses a heterogeneous catalyst adopting a hierarchical pore structure and adopting phosphine-containing organic mixed polymer-metal as well as preparation and an application of the heterogeneous catalyst in production of aldehyde with a high normal/isomeric ratio through internal olefin hydroformylation. In the heterogeneous catalyst, one, two or three of metal Rh, Co and Ir are taken as active components, and a phosphine-containing organic mixed polymer adopting the hierarchical pore structure is taken as a carrier and is prepared from multidentate organic phosphine ligands containing alkylene and monodentate organic phosphine ligands through copolymerization. The coordinate bond type heterogeneous catalyst is applicable to a fixed bed reactor, a slurry bed reactor, a bubble column reactor, a trickle-bed reactor and other reactors. The provided heterogeneous catalyst has good presentation in the internal olefin hydroformylation reaction, internal olefins are firstly isomerized to be terminal-group olefins under the action of the catalyst, a hydroformylation reaction is performed, corresponding normal aldehyde is selectively generated, accordingly, the normal/isomeric ratio of the product aldehyde is high and can be higher than 15, and the content of alkane in the obtained product is lower than 1%.

The invention discloses an image enhancement method in a dark light environment and belongs to the field of image processing. The problems of distorted representation result, blurred image contour andlow color saturation after enhancement seriously exist can be solved; The method comprises the following steps: extracting a brightness component V in an HSV in a dark light image, carrying out logarithmic transformation of global brightness, and carrying out overall brightness improvement on a low-illumination area in the dark light image; Obtaining neighborhood average brightness of pixel points in the current dark light image by using bilateral filtering under a receptive field three-Gaussian model of the retinal neurons; A function of extracting a local bright part from an image background is realized by utilizing top and bottom cap transformation processing, and then light and shade extraction and noise reduction are carried out on the image by utilizing top and bottom cap transformation; According to the method, the enhanced details of the dark-light color image can be well represented, the problem of non-uniform illumination distribution is solved, the overall contrast of the image is improved, the detail contour of the image is clear, and the color saturation is high.

The invention relates to an online transaction fraud detection method based on an entity relationship, which is characterized in that the entity relationship is extracted according to transaction data, a relational network bipartite graph is constructed, and a heterogeneous network homogenization method based on node contraction and a neighborhood information aggregation promotion tree classification model mechanism based on ensemble learning and graph representation learning are provided. According to the method provided by the invention, from the perspective of practicability, the attentionis converted from the transaction node to the multi-order neighborhood information of the transaction in the relational network, and the potential association relationship between the transactions isfully considered, so that the possibility is provided for mining gang fraud. The gradient promotion model improves the fraud identification effect through continuously fitting the residual error of the model, and has a very good performance effect. Meanwhile, the method expands the ensemble learning from the application of the grid-type data to the application field of the graph data. Based on theabove aspects, a framework of the loan transaction fraud detection method is established, and technical support is provided for solving fraud transaction detection.

The invention discloses a hypervideo construction method based on rough drawings, which belongs to the technical field of computer application. The method of the invention comprises the following steps: establishing the scene rough drawings of each target video, and establishing the recessive semantic association relationship between the rough drawings and the target video; 2) searching the target video resources according to the established scene rough drawings, respectively adding noted rough drawings into each target video resource, and establishing the dominant semantic association relationship between different video resources; and 3) setting the behavior attribute of the noted rough drawings, establishing the association relationship between the target video expressed by the current rough drawing and other target video sources, and obtaining the hypervideo of the target video resources. Compared with the prior art, the invention has the advantages that the constructed hypervideo conforms to the cognizancing habits of people, can perfectly express the complicated semantic association between the videos, and simultaneously breaks through the traditional single linear passive viewing method.

The invention discloses a method for high selectively preparing an aldehyde by using an olefin. According to the method, an organic polymer self-load type catalyst is used, an olefin hydroformylationreaction is applied to high selectively preparation of an aldehyde. According to the organic polymer self-load catalyst, one, two or three metals selected from a group consisting of Rh, Co, Pd and Irare used as active components, and a phosphine-containing organic copolymer is used as a carrier, wherein the phosphine-containing organic copolymer carrier is formed by copolymerization of a polydentate organic phosphine ligand and a monodentateorganic phosphine ligand. The organic polymer self-load catalyst provided by the invention is suitable for reactors such as fixed beds, slurry beds, bubbling beds, trickle beds and the like, and has good performance in a reaction of high selectively preparing the aldehyde through the olefin hydroformylation reaction; the aldehyde with a high normal-specific ratio can be produced, wherein the positive-specific ratio can reach 25 or above; side reactions are few; and the alkane content of an obtained product is 1% or below. The multi-phase catalyst is good in stability, and the separation of the catalyst from the reactants and products is simple and efficient.

The invention discloses a voice activity detection method and equipment based on a time domain and a frequency domain. The method comprises the steps that noise energy of voice signals is estimated; white noise is added into the input voice signals; the voice signals after the white noise is added are processed in a framing mode; a short-time energy value of each frame is determined; a harmonic wave product spectrum value of each frame is determined; for each frame, whether the frame is a voice frame is determined according to the short-time energy value and the harmonic wave product spectrum value of the frame, and voice segments contained in the voice signals are obtained. According to the voice activity detection method and equipment, a time domain analysis method and a frequency domain analysis method are combined, the adaptability is high for the actual condition, performance is good under the circumstances that the background noise is human voice and the signal to noise ratio is low, the method is easy and ingenious to realize, and the equipment can be embedded into various voice systems easily.

The invention discloses a phosphine-containing organic mixed polymer-metal heterogeneous catalyst and its use in butyraldehyde production. The heterogeneous catalyst comprises one, two or three of Rh, Co and Ir as active ingredients and utilizes a multistage pore-structure phosphine-containing organic mixed polymer as a carrier. The phosphine-containing organic mixed polymer is prepared from an alkylene-containing polydentate organic phosphine ligand and a single dentate organic phosphine ligand through copolymerization. The coordination bond-type heterogeneous catalyst is suitable for a fixed bed reactor, a slurry bed reactor, a bubbling bed reactor and a trickle-bed reactor. The heterogeneous catalyst has good effects in hydrogen-propene formylation preparation of butyraldehyde, can produce high-normal/isomeric ratio butyraldehyde with a normal/isomeric ratio of 25 or more and realizes product propane content less than 0.1%. The heterogeneous catalyst has good stability and can be simply and efficiently separated from reactants and products.

The invention provides an animation-based action prediction generation method and device, wherein the method comprises the steps of obtaining a current animation frame of a target animation corresponding to a target role; obtaining the current skeleton posture information of the target character and the motion posture information on a preset motion track according to the current animation frame; acquiring a similar action corresponding to the current animation frame and a similarity corresponding to the similar action; obtaining a fusion action according to the similarity and the similar action; obtaining a prediction action at the next moment; and controlling the target animation to simulate the prediction action in the next animation frame. According to the invention, the network learning is carried out on the motion characteristics of an action body with a skeleton structure, such as a human, an animal, etc., in different motion states and the change characteristics between the motion states; and finally, different motion characteristics can be well represented, and the change between different motion states can be naturally fused, so that the advancing animation effect of the people or animals is produced, and the natural effect of the animation effect is ensured.

The invention discloses an image description generation method based on an architectural short sentence constraint vector and a dual visual attention mechanism, A semantic model for automatically describing the visual content information of images is obtained by training a large number of labeled texts. The semantic model is composed of three parts: a short sentence generation model, a dual visualattention mechanism and a constrained language model. The text description can be generated automatically for any input test image. The invention can effectively establish the relationship between the word and the image in the text description, and has good performance for describing the salient objects or scenes of the image based on semantics.

The invention discloses a sea ice calculating method of using a marine navigation simulator to simulate a ship-ice interaction process. The sea ice calculation is divided into an ice breaking calculation module and a rigid body motion module according to current sea condition information, and sea ice force, sea ice brash ice shapes, and sea ice motion states during an ice breaking process are calculated and analyzed, and then generated results are transmitted to a ship motion model and an ice area visual system. The sea ice calculating method is used to fill in the blank of the sea ice calculating method of the marine navigation simulator, and a semi-infinite plane elastic foundation theory, a tapered beam structure, and an energy conservation theory are used in the corresponding modules, and then brash ice forming conditions, brash ice bending stress, and brash ice energy distribution and motion states are calculated. Sea ice physical characteristics are well displayed in the ice area simulation visual system of the marine navigation simulator, and the sea ice calculating method is used as a physical foundation of ship-ice interaction process visualization, the ice breaking process is dynamically and vividly simulated, and then reliable sea ice stress feedback is provided for a ship motion model.

The invention relates to a production method and applications of cyclic carbonate. According to the present invention, the used catalyst is a multifunctional biomimetic heterogeneous catalyst formed by an organic porous copolymer based on metalloporphyrin and a quaternary phosphonium salt, the catalyst carrier is formed by carrying out mixed polymerization on alkylene-functionalized metalloporphyrin and a quaternary phosphonium salt, and the active component of the catalyst comprises metalloporphyrin (as a Lewis acid activation epoxy compound) and a quaternary phosphonium salt (as a ring-opening nucleophilic reagent) functional group; the obtained multifunctional biomimetic heterogeneous catalyst can be used for fixed beds, slurry beds, kettle type reactors, trickle beds and other reactors; the obtained multifunctional biomimetic heterogeneous catalyst has good performance, extremely high activity (2000-6000 h<-1>), good selectivity (more than 99%) of the formed cyclic carbonate and strong substrate applicability in the cyclic carbonate production reaction between the epoxy compound and the carbon dioxide, further has good stability, and can be simply and efficiently separated fromthe substrate and the product.

The invention provides a cross-age face recognition method. According to the method, a cross-age face recognition system composed of two modules (a maximum entropy feature description module and an aging perception de-noising automatic coding module) is obtained by training a large amount of face images including four age groups, thereby realizing recognition on any two different age face images. The maximum entropy feature description module allocates codes including maximum information amount by using maximum entropy splitting of a decision tree, the aging perception de-noising automatic coding module reconstructs a feature descriptor of any age group into feature descriptors of four different age groups, the descriptors are integrated to obtain a face integrated feature vector for eliminating aging influencing, and face recognition is finally realized by calculating the cosine distance of integrated feature vectors of different faces. The method can well reduce the information loss problem of some traditional descriptors, eliminates the influence of the aging factor in cross-age face recognition, and has good performance in the cross-age face recognition problem.

The invention discloses a surface carving technology of metal jewelry. The technology comprises the following steps: (1) processing raw materials: melting and then drawing a metal into a tabular shape; punching the tabular metal into a corresponding shape; and then cutting and trimming the metal; (2) setting parameters and compiling processing data: compiling words or patterns preliminarily designed into the processing data; (3) cutting and carving: cutting and carving the words or the patterns on the cut deformed metal by layers according to the processing data; and (4) subsequently processing: grinding and smoothening the metal with the carved words or the patterns. Patterns carved on the surfaces of metal ornaments can be clear and obvious by the technology to have smooth lines and favorably present the contents of the patterns so as to improve the ornament and the artistry.

The invention discloses a time sensitive transmission and bandwidth optimization utilization-based method used under an IEEE802.15.4 network environment, which comprises the following steps of: firstly, considering a time period Di of transmission data of different nodes and time ED (i) occupied by transmission; and secondly, allocating GTS for each node, reallocating GTS (UGTS) which is not usedin the last scheduling process by an effective method, and improving the GTS utilization rate so as to improve the bandwidth utilization rate. The method has the advantages of good performance in theprocess of transmitting time sensitive service and convenient implementation at a user terminal.

The invention discloses an intelligent internal rearview mirror real-time deblurring method based on a neural network algorithm. The invention collects the road traffic video to generate blurred frameand clear frame, then repeatedly trains the CNN deblurring neural network with time-space recursion to obtain the optimal deblurring model, then takes the video image of the rear pavement collected by the camera as the input, and the obtained output is the dynamic video after deblurring processing. The invention provides an intelligent inner rearview mirror realization replacing the traditional outer rearview mirror, which can reduce the wind resistance, eliminate the visual blind area of a large freight car, and is suitable for bad weather and the like. Compared with the traditional image processing method, the network generated by depth neural network has better effect of deblurring. Improve the traditional CNN deblurring architecture, put forward the idea of space-time recursion, without increasing network parameters to expand the network acceptance domain, so that the efficiency of deblurring is greatly improved, can achieve real-time, and achieve the same effect as the traditional rearview mirror.

The present invention relates to a video content visual analysis method based on a map metaphor, and belongs to the technical field of computer application. The method comprises: firstly, performing structuring processing according to an original video content to obtain data in a corresponding structure, and establishing a semantic association relationship between a map and a target video; then, performing layout algorithm processing on the processed data, so as to obtain a basic layout of a visualization system, and according to a mapping rule of a map metaphor system, mapping a video content and association onto a video map to generate the video map visualization system; and then implementing a draft interactive interface and other related functions such as counting, analyzing and querying paths. Compared with the prior art, the video map constructed by the method provided by the present invention meets cognitive habits of people, and based on an advantage of map expression, excellently presents a complex semantic relationship between the videos, and meanwhile breaks a conventional single linear passively browsing mode, and helps users browse the videos better.

The invention relates to a target reconstruction method based on dictionary learning. Firstly, existing dense point cloud models are utilized for constructing a point cloud dictionary library; then asparse point cloud model of the target is constructed, and is expanded by virtue of the point cloud dictionary library for obtaining a complete dense three-dimensional model, and features are constructed in a process of expansion on the basis of local curvature invariance of point cloud surface pieces to be used as bases of expansion; and finally, surface reconstruction is carried out on the model, which is obtained by expansion in the previous step, to complete target reconstruction. The method can greatly reduce computation time, and has very good performance for reconstruction of targets ofwhich image texture is not rich or texture areas are repeated.

A method for synthesizing human face expression based on dense feature correspondence and morphology includes realizing effective separation of position information and grey scale information of human face feature in image by using decomposition-presentation of human face image and arresting shape and vein difference in expression variation with no interference to each other, utilizing morphological means to pick up black-white spot noise region and mapping said difference on new human face image as well as using linear interpolation mode to filter out noise for finalizing expression synthesis.

The invention relates to a triphenylamine containing isoindigo polymer, a preparation method of the triphenylamine containing isoindigo polymer, and an application of the triphenylamine containing isoindigo polymer to electrochromism, and relates to an isoindigo polymer and a preparation method and an application of the isoindigo polymer. The invention aims at solving the problem of relatively narrow application range of isoindigo caused by the low solubleness of the isoindigo in an organic solvent. The structural formula of the triphenylamine containing isoindigo polymer is shown in the description, wherein n in the formula is a positive integer. The preparation method is as follows: 1, synthesizing N<1>,N<1>-bi(4-bromobenzene group)-N<4>,N<4>-diphenylbenzene-1,4-diamine monomer; 2, preparing an isoindigo derivative; 3, preparing the triphenylamine containing isoindigo polymer. The invention discloses the application of the triphenylamine containing isoindigo polymer to electrochromism as an electrochromic layer in an electrochromic device. The polymer has the electrochromic property and the memory property, can be applied to the electrochromism field, and is excellent in explosive detection and photoelectric detection. The triphenylamine containing isoindigo polymer is suitable for preparing the isoindigo polymer and the application of the isoindigo polymer.

The invention provides an analog circuit intermittent fault diagnosis method based on a multi-granularity cascade forest. In order to solve the problem that the traditional fault diagnosis method is not suitable for large-scale complex integrated circuits at the present stage, and the tntermittent fault data is small, the characteristic that a multi-granularity cascade forest utilization algorithmis adopted to well express small data is adopted to classify different types of fault data so as to diagnose the intermittent faults of an analog circuit. According to the method, a deep neural network layer-by-layer cascade structure is adopted; the depth of a cascade layer and the number of random forests and complete random forests in each layer can be adjusted according to actual conditions,small-scale intermittent fault data can be trained to achieve a good result, the requirement of deep learning for a large-scale data set is avoided, and meanwhile the detection effect comparable to that of a deep learning algorithm can be achieved.

The invention discloses a track-energy-diffusion-diagram-based group behavior identification method. The track-energy-diffusion-diagram-based group behavior identification method comprises the following steps: 1, carrying out multi-target detection and multi-target tracking on an original video and extracting a multi-target track; 2, converting the multi-target track into a track energy block diagram; 3, generating a track energy diffusion diagram by the track energy block diagram by virtue of an energy diffusion process; 4, generating a track energy equipotential line diagram by the track energy diffusion diagram; 5, carrying out feature extraction of the track energy equipotential line diagram, thereby obtaining fusion feature vectors by a canonical-correlation-analysis-based feature fusion algorithm; and 6, classifying and identifying the fusion feature vectors. According to the track-energy-diffusion-diagram-based group behavior identification method, the track energy block diagram has a high performance and is capable of saving the time information of the group behavior track; the track energy diffusion diagram is capable of smoothening noise randomly generated by movement; the multi-feature fusion is implemented by the canonical-correlation-analysis-based feature fusion algorithm, so that the identification accuracy is improved.