71 results about "Simple Features" patented technology

The invention discloses a modulation mode identification system and method based on a convolutional neural network, which solve the problems of complex feature extraction steps and low identificationrate under a low signal-to-noise ratio in the prior art. The simple feature in the identification system is constructed as a simple feature using a co-directional component and a quadrature componentof a baseband signal as signals, and the simple feature is sent to a convolutional neural network module for identification. The identification method comprises the steps of: modulating a transmittedsignal and performing pulse shaping; performing up-conversion on the transmitted signal and then transmitting the transmitted signal through an additive white Gaussian noise channel; performing pre-processing first by a receiving end to obtain the co-directional component r(t) of the analyzed signal; constructing the simple feature, i.e., constructing the co-directional component r(t) and the quadrature component of the analyzed signal into a two-dimensional matrix; performing feature learning and classification by the convolutional neural network; and sending a modulation method to a demodulation end to obtain a demodulated signal. The method is low in feature design complexity, avoids explicit feature extraction, has high classification correctness, and can be applied to communication systems having high recognition performance requirements.

The present invention provides a deep convolution neural network-based traffic flow density estimation method. The method comprises the steps of collecting road video images by a camera, sending multi-scale pyramid image blocks into a convolution neural network through image pre-processing, extracting simple features and abstract features from a bottom layer to a high layer, and obtaining a distribution density image of each scale of traffic flow image; learning a mapping from the multi-scale distribution density image to a distribution density image of an overall image and a number of all vehicles in the image by using a fully connected network layer; dividing regions of interest of a distribution density image of a video image output by the convolution neural network, summing pixels of the regions of interest so as to obtain a number of vehicles of a single lane or multiple lanes; and calculating according to a region length so as to obtain an instantaneous traffic flow density of the region. Through adoption of the method, vehicles are calculated and traffic flow density is estimated in real time more accurately.

The present invention provides an automatic fastening and loosing shoes, which is typically comprised of a shoe body including tongue portion, upper portions and shoe-sole, in which mouthing portions are built on the interfaces of the tongue portion and the upper, said shoe-sole includes an outsole, a middle sole and a bottom filler; said bottom filler is fastened on the middle sole with the front portion so that the middle and heel portions can took off from the middle sole; an action regulating mechanism includes a flutter foil, a pair of guide plates and two inserting strips; said flutter foil is connected to the both sides of the tongue with the both upper ends respectively, cooperating to the guide assembly combined by the guide plates and the inserting strips pre-setting in upper-low position, and connected to the bottom filler at the rear portion with the another end, thereby towing the tongue can bring to the heel portion of the bottom filler simultaneously and relatively moving relative to the outsole cooperating to the attempting action; a control mechanism is consisted of a switch, a hook plate, a stationary guide and a snap frame, said hook plate and said stationary guide are attached on the heel portion of the bottom filler for locking on said snap frame located on the middle sole cooperating to put-on attempting action, and the lock of the hook and the snap frame is released by the switch mounted on the rear end of the shoe-body for carrying out automatic put-off function. The structure has less components and simple features without tying and loosing lace, so that the action of the structure can be carried out smoothly and surely, with low rejection rate and low production cost.

The aesthetic quality of a picture is automatically inferred using visual content as a machine learning problem using, for example, a peer-rated, on-line photo sharing Website as data source. Certain visual features of images are extracted based on the intuition that they can discriminate between aesthetically pleasing and displeasing images. A one-dimensional support vector machine is used to identify features that have noticeable correlation with the community-based aesthetics ratings. Automated classifiers are constructed using the support vector machines and classification trees, with a simple feature selection heuristic being applied to eliminate irrelevant features. Linear regression on polynomial terms of the features is also applied to infer numerical aesthetics ratings.

The present approach relates, in some aspects, to a multi-level and a multi-channel frame work for segmentation using model-based or “shallow” classification (i.e. learning processes such as linear regression, clustering, support vector machines, and so forth) followed by deep learning. This framework starts with a very low resolution version of the multi-channel data and constructs an shallow classifier with simple features to generate a coarser level tissue mask that in turn is used to crop patches from the high-resolution volume. The cropped volume is then processed using the trained convolution network to perform a deep learning based segmentation within the slices.

An algorithm for the detection of military vehicles in FLIR imagery is used as a prescreener to eliminate large areas of the image from further analysis. The output is a list of likely target locations with confidence numbers, which may be sent to a more complex clutter rejection algorithm for analysis. The algorithm uses simple features, and is intended to be applicable to a wide variety of target-sensor geometries, sensor configurations, and applications.

The aesthetic quality of a picture is automatically inferred using visual content as a machine learning problem using, for example, a peer-rated, on-line photo sharing Website as data source. Certain visual features of images are extracted based on the intuition that they can discriminate between aesthetically pleasing and displeasing images. A one-dimensional support vector machine is used to identify features that have noticeable correlation with the community-based aesthetics ratings. Automated classifiers are constructed using the support vector machines and classification trees, with a simple feature selection heuristic being applied to eliminate irrelevant features. Linear regression on polynomial terms of the features is also applied to infer numerical aesthetics ratings.

The embodiment of the invention discloses a video classification method and device, a model training method and device, equipment and a storage medium, and belongs to the technical field of computer vision. The method comprises the following steps: acquiring a video; selecting n image frames from the video; extracting respective feature information of the n image frames through a feature extraction network according to a learned feature fusion strategy; and determining a video classification result according to respective feature information of the n image frames. According to the learned feature fusion strategy, feature information of the image frame is extracted; wherein the feature fusion strategy indicates that each image frame is fused with feature information of other image frames; the proportion of the feature information of each image frame is determined, the classification result of the video is determined according to the feature information of the image frames, the feature fusion strategy replaces complex and repeated 3D convolution operation through simple feature information fusion, the workload is small, the time for finally obtaining the classification result of thevideo is short, and the efficiency is high.

The invention discloses a single-point multi-view meter-hanging posture error detecting method of a mobile industrial robot. The single-point multi-view meter-hanging posture error detecting method includes firstly determining position error of an end effector from a demonstration target point under a tool coordinate system initially by photographing a feature point, moving the end effector by a certain distance under the tool coordinate system along the coordinate axis direction, photographing and matching and comparing the picture with the last picture by templates, determining relative error of the feature point and finally determining position and posture error of the end effector. The single-point multi-view meter-hanging posture error detecting method mainly adopts algorithm to determine posture error the tail end of the moving industrial robot in the plane vision positioning and can determine through only one simple feature point, thus, the problem of limitations (such as camera visual angle, focal length and cost) in actual application is solved, image processing efficiency and speed (single-point processing) are improved to some extent, and the problem of instantaneity in image processing in the industry is solved.

The invention discloses a method for target correlation and track generation of an image sequence. The method comprises the steps that a multistage correlative mechanism is designed, multiple static features and dynamic features of targets are extracted stage by stage, and different target affinity distance models are formed on different correlation stages, so that the multiple features of the targets are fully comprehended to conduct data correlation, and the accuracy of correlation is improved; meanwhile, based on a data correlation result, in the same image background, matched target center points are connected stage by stage, each obtained track section expands to the two ends continuously, and finally a complete motion track of each target is obtained. According to the method, various simple features of the targets are extracted stage by stage, so that multi-target correlation and track generation of a single-camera video are achieved reliably and rapidly, and the calculated amount is extremely small.

The invention provides a common-path self-calibration film thickness measuring device and a measuring method based on polarization multiplexing. The measuring device includes a light source output module, a film thickness measuring probe module, a demodulation interferometer module, a polarization beam splitter module, and an acquisition and control module. The polarization multiplexing technology is used in the invention, and two probes use orthogonal polarized light. The measuring probes can transmit and reflect transferred light. When there is no film to be measured, the absolute distance H between the two probes can be measured. When a film to be measured is placed between the two probes, the absolute distances H1 and H2 between the two probes and the front and back surfaces of the film to be measured can be measured. The thickness d of the film to be measured is determined according to the formula d=H-(H1+H2). The thickness of a film to be measured can be measured without a calibration object. Through a common-path design, the influence of the mechanical instability inside the system and the change in external environment on the measurement process is overcome. The measuring device and the measuring method have the advantages of self-calibration, simple feature white light interference peak identification, large dynamic range, traceable measurement result, and the like.

The aesthetic quality of a picture is automatically inferred using visual content as a machine learning problem using, for example, a peer-rated, on-line photo sharing Website as data source. Certain visual features of images are extracted based on the intuition that they can discriminate between aesthetically pleasing and displeasing images. A one-dimensional support vector machine is used to identify features that have noticeable correlation with the community-based aesthetics ratings. Automated classifiers are constructed using the support vector machines and classification trees, with a simple feature selection heuristic being applied to eliminate irrelevant features. Linear regression on polynomial terms of the features is also applied to infer numerical aesthetics ratings.

This invention relates to the fabrication of large area nanoimprint molds having complex patterns with minimal or no use of direct-writing, such as electron beam lithography, ion, laser beam, or mechanical beam lithography. This can be accomplished by forming a pattern of simple nanoscale features and converting the simple features into more complex nanoscale features by a process comprising shadow deposition. The process may also include steps of uniform deposition, etching and smoothing depending on the shape of the complex features.

The invention discloses a behavior recognition method based on a space-time attention enhancement feature fusion network, and belongs to the field of machine vision. According to the method, a networkarchitecture based on an appearance flow and motion flow double-flow network is adopted, and is called as a space-time attention enhancement feature fusion network. Aiming at a traditional double-flow network, simple feature or score fusion is adopted for different branches, an attention-enhanced multi-layer feature fusion flow is constructed to serve as a third branch to supplement a double-flowstructure. Meanwhile, aiming at the problem that the traditional deep network neglects modeling of the channel characteristics and cannot fully utilize the mutual relation between the channels, the channel attention modules of different levels are introduced to establish the mutual relation between the channels to enhance the expression capability of the channel characteristics. In addition, thetime sequence information plays an important role in segmentation fusion, and the representativeness of important time sequence features is enhanced by performing time sequence modeling on the frame sequence. Finally, the classification scores of different branches are subjected to weighted fusion.

The invention relates to an indoor autonomous positioning method of a mobile robot which combines the point and line characteristics of a scene. The method has strong scene adaptability, which uses adepth camera mounted on a mobile robot to collect scene video data, and extracts the point and line features of each frame of the scene to complete the subsequent positioning calculation. Generally, point features perform well in clear texture, rich feature points and unobstructed scenes. However, in the scene with fewer texture feature points, there is a problem of scarcity of feature points, andthe robustness of the simple feature point location method is poor. The invention extracts more stable object line features in the scene with missing texture, thereby ensuring to provide more abundant feature information to realize the calculation of the visual odometer, thereby obtaining the position information of the mobile robot and achieving the purpose of autonomous positioning.

The invention discloses a DNA quantitative analysis method based on a cell microscope image. The method comprises the following steps that 1) pretreatment is carried out; 2) an MSER algorithm is used to extract candidate areas, and single-tine trees or single nodes are obtained; 3) a simple feature training classifier C1 is extracted from the area on the basis of a training set, the possibility that the classifier belongs to a cell of interest is calculated, and the single-tine trees obtained in the step 2) are further simplified based on the classifier; 4) a statistical, shape and texture feature training classifier C2 are extracted on the basis of simple features of the extraction area of the training set, and the classifier C2 is used to further classify all the candidate areas of positive samples output by the classifier C1; and 5) a DI value of a nucleus is calculated. According to the invention, a maximum stable area algorithm is used to extract a lot of candidate areas from the image, and a cascaded classifier composed of the classifier C1 and the classifier C2 is trained to identify the candidate areas rapidly and accurately.

The invention discloses an automatic malicious code variation detection method. The method comprises the following steps of: feeding a to-be-detected program into a program quality detection system, so as to detect whether the to-be-detected program is a strictly executable program or not: verifying simple features such as a size, a suffix and a naming rule of the program; storing the to-be-detected program which passes the program quality detection system in a database, and handing over the to-be-detected program to a malicious code variation detection system to carry out detection; and if the quantity of newly detected malicious codes stored in a database system achieves a preset threshold value, triggering a set trigger in the database system, so as to enable an automatic fine tuning system to start work. According to the method, a software component technology is flexibly applied to design and development of systems, so that the reliability and extensibility and maintainability ofthe systems are improved, the malicious code variation detection system perfect in function is realized, and the current manner of manually selecting malicious code to detect features is improved.

A simplified television remote control device with limited control features for on / off, channel selection, and volume. Volume and channel control are provided as relatively large luminescent knobs which can be turned clockwise or counterclockwise. The housing is tapered so that its pointing direction can be determined by feel. The device self-learns the limited control features from an old remote control device for the television by placing the old device nose-to-nose with the simplified device and pressing a specific sequence of controls on the old device. A photo diode senses the infrared command signals from the old device and associates those commands with the control features of the simplified device.

A kind of 2D bar reader includes image reading head, terminal of computing device. The said image reading head through control interface circuit is connected to terminal of computing device with engine software for reading and recognizing bar code being installed on. The said terminal of computing device is one of desktop computer, server, palm computer, embedded type system or handset possible to install engine software for reading and recognizing bar code. It is initiation of the invention that recognition function is transferred to terminal of computing device connected. Thus, the image reading head possesses features of simple feature, easy of manufacture, high processing capacity, good recognition effect.

A device and method for preservation of an oxidizable substance, e.g. a liquid such as a beverage of which wine is an example, is described. In a first aspect, a device is described for controlling the delivery of a predetermined amount of inert gas into a container with a single opening such as a wine bottle at a pre-determined pressure, suitable for an efficient delivery of the inert gas, with a simultaneous displacement of the air present in the top space of the container or bottle, in a simple actuation. In a second aspect, an improved container or bottle stopper is described with simple features that allow easy and cheap manufacture significantly reducing the cost of the stopper. In a third aspect to an efficient and cost-effective method is described for the preservation of a degradable substances especially liquids such as beverages of which wine is an example.

The invention discloses a feature compensation video target tracking method based on a posterior pixel color histogram, a direction gradient histogram and a convolutional neural network, which uses simple features in a simple scene to ensure real-time performance, and uses complex features in a complex scene to ensure accuracy. Through combination of two features of a posterior pixel histogram anda direction gradient histogram, an obtained response feature map can well adapt to a simple condition of a video scene. Training a classifier to judge which situation the response obtained by the former is fused is untrusted; and finally, according to the judgment result of the classifier, selecting whether to add a convolutional neural network tracker which is slow in speed and more robust in performance to correct the target deviated from the tracking, or re-finding the target lost from the tracking. According to the method, the precision of judging the size and the position of the target in the video is improved, and the method can well adapt to a long-time target tracking task to achieve a scene of practical application.

The invention relates to a rapid detection method for a high-resolution remote-sensing image airport area-of-interest. Firstly, an LSD algorithm and an FT algorithm are adopted to calculate a linear detection binary result map and a saliency map of a high-resolution remote-sensing image respectively, then a huge number of rectangular windows are elected randomly in the high-resolution remote-sensing image, and then object metrical scores of the rectangular windows are calculated, and then a small number of rectangular windows are selected to form the airport area-of-interest through adoption of a non-maximum suppression algorithm according to the object metrical scores of the rectangular windows. The rapid detection method for the high-resolution remote-sensing image airport area-of-interest makes use of complementarity of two simple features and is capable of detecting the area-of-interest stably with a small calculation quantity so that high-resolution remote-sensing airport detection speed and precision are greatly improved. Therefore, the method is high in practicality.

The invention discloses a decomposition method of image texture and structure based on a total variational model of a guide map, belonging to the technical field of image processing. According to theinvention, firstly, the local main structure of the image is reconstructed based on the guide image filter, then the texture descriptor is calculated according to the reconstructed local structure image, finally, the accuracy and computational efficiency of texture and structure decomposition are improved by combining the multi-scale total variational model and block translation method. The technical scheme of the invention can obtain better texture / structure decomposition effect for noisy images, the textures and structures of different scales can also be decomposed accurately, and the decomposed structure layers can keep the original shading of the image, avoiding the blurring of the structure caused by local smoothing or the color block effect caused by the global optimization method. In addition, the technical proposal of the invention needs to extract simple features, and does not depend on learning a large number of image samples.

The invention provides a sample block-based image target counting method. The method comprises steps: image blocks with a fixed size are sequentially extracted from an input image through a sliding window; and then, according to the simple features and a similarity measurement function, the most similar K candidate image blocks are searched from a training set. Based on the K blocks, sparsity constraints are used, few samples for reconstruction are selected, and reconstruction weights corresponding to the samples are calculated. The weights are applied to a density map corresponding to the samples, an extracted image block corresponding density map is obtained, and the density map is placed at a corresponding position on an input image density map. The above process is repeated until all image blocks are extracted through the sliding window. Finally, all pixel values in the input image density map are accumulated to obtain the number of targets of interest. Compared with a mainstream method, the method of the invention has the advantages that the needed training images are few, the features are simple, the satisfactory accuracy can be achieved, the image resolution is robust, and the good counting accuracy can be kept even if the input image or a video stream has low resolution.

A device and method for preservation of an oxidizable substance, e.g. a liquid such as a beverage of which wine is an example, is described. In a first aspect, a device is described for controlling the delivery of a predetermined amount of inert gas into a container with a single opening such as a wine bottle at a pre-determined pressure, suitable for an efficient delivery of the inert gas, with a simultaneous displacement of the air present in the top space of the container or bottle, in a simple actuation. In a second aspect, an improved container or bottle stopper is described with simple features that allow easy and cheap manufacture significantly reducing the cost of the stopper. In a third aspect to an efficient and cost-effective method is described for the preservation of a degradable substances especially liquids such as beverages of which wine is an example.

The present approach relates, in some aspects, to a multi-level and a multi-channel frame work for segmentation using model-based or “shallow” classification (i.e. learning processes such as linear regression, clustering, support vector machines, and so forth) followed by deep learning. This framework starts with a very low resolution version of the multi-channel data and constructs an shallow classifier with simple features to generate a coarser level tissue mask that in turn is used to crop patches from the high-resolution volume. The cropped volume is then processed using the trained convolution network to perform a deep learning based segmentation within the slices.

The invention relates to the field of GISs, in particular to a GIS point buffer area analysis method and system. The method and system are characterized in that the method comprises the steps that point map layer data which is about to be subjected to spatial buffer analysis is read in batch, and a corresponding memory structure is built; multi-thread parallel buffer analysis is conducted on all point features read by a system memory; data information of the point features which are subjected to buffer analysis is read and stored in a point simple feature-class map layer of a corresponding database in batch, and then a buffer analysis result is obtained; the system comprises a data read unit, a data buffer unit and a data storage unit. According to the method and system, a multi-thread parallel buffer technology is adopted, and high efficiency and stability of GIS point buffer area analysis are achieved.

The invention discloses a back-bolt type stone curtain wall which comprises a cross beam, wherein a connecting part is firmly connected to the cross beam; the connecting part has an L-shaped structure; the upper end of the connecting part is inserted in a hanging part; and the hanging part is connected to a stone material by a back bolt. The back-bolt type stone curtain wall is characterized by convenient and easy installation; panels of the stone material are not damaged; and the structure is simple and easy-to-achieve.

The invention discloses a robot vision control method based on the rectangle feature. The robot vision control method comprises the steps that taking a rectangle as the visual feature, visual controlis conducted through vertexes, the area and included angles of diagonal lines of the rectangle; according to the deviation between the current coordinates of the vertexes of the rectangle and the expected coordinates, the position deviation of the robot is obtained, according to the deviation between the current area of the rectangle and the expected area, as well as the deviation between the current included angles and the expected included angles of the diagonal lines, the attitude deviation of the robot is obtained, and finally six-degree-of-freedom control of the robot is achieved; and translation motion and rotational motion are decoupled, and the control process is more stable and more reliable. The problem that traditional visual control is prone to being affected by the shape, thetexture, noise, the lighting condition and the like of a target object due to the fact that the simple features such as points, edges and straight lines are extracted is solved.