52results about How to "Feature stabilization" patented technology

A web transfer belt is provided having a good paper releasability with good paper releasability and having stable surface features. The web transfer belt includes a high-molecular weight elastic member 2 having a web-receiving face 2b for receiving a web P to transfer, and a surface layer forming member 3 disposed on the high-molecular weight elastic weight member 2. A portion of high-molecular weight elastic weight member 2 is exposed on web-receiving face 2b, with one of the web-receiving face 2b and surface layer forming member 3 being made of a hydrophobic material. In this way, water-repellent portions and water-condensing portions are dispersed on the web-receiving face, with recesses and protrusions at fine scales, which can break a water film formed between web-receiving face 2b and web P, and can be easily formed on the web-receiving face 2b.

The invention discloses a high-spatial resolution remote-sensing image bag-of-word classification method based on linear words, which includes first dividing images to be classified into a practice sample and a classification sample. Steps for the practice sample include collecting linear characteristics of the practice image and calculating linear characteristic vector; utilizing K-Means++ arithmetic to generate linear vision word list in cluster mode; segmenting practice images and obtaining linear vision word list column diagram of each segmentation spot block on the base; and conducting class label on the spot block and putting the classification and linear vision word column diagram in storage. After sample practice, steps for the classification sample include collecting linear characteristics of the images to be classified, segmenting the images to be classified, calculating linear characteristics vector on the base, obtaining linear vision word list column diagram of each segmentation spot block and selecting an SVM classifier to classify the images to be classified to obtain classification results. The high-spatial resolution remote-sensing image bag-of-word classificationmethod utilizes linear characteristics to establish bag-of-word models and is capable of obtaining better high spatial resolution remote sensing image classification effect.

An electric power transmission system includes: a transmitting-side system that includes switching elements (SW1 and SW2) that convert a direct-current voltage to an alternating-current voltage and that output the alternating-current voltage and a transmitting-side magnetic resonance antenna unit (120) that has a first inductor (121) and a first capacitor (122) directly coupled to each other and to which the output alternating-current voltage is input; and a receiving-side system that includes a second inductor (221) and a second capacitor (222) directly coupled to each other and that resonates with the transmitting-side magnetic resonance antenna unit (120) via electromagnetic field to thereby receive electric energy output from the transmitting-side magnetic resonance antenna unit (120).

The invention belongs to the technical field of video detection in deep learning. The invention relates to a vehicle detection method, in particular to a rapid vehicle detection method based on windowcharacteristics. According to the method, a vehicle window is used for replacing a vehicle body to serve as a target object for detection, a residual module of a ResNet network and a multi-scale feature extraction method of an SSD algorithm are combined, a network structure of YOLOv3 is used for reference, and a full convolution detection method with only 24 convolution layers is constructed; Under the condition that the traffic flow is large, during batch testing, the average detection precision is close to 100%, the average detection rate reaches 90%, the detection speed reaches 22 milliseconds per frame, real-time detection of vehicles in a road high-definition monitoring video is achieved, the detection rate of the vehicles in the large traffic flow is effectively increased, and the method has important application value.

The invention is the recognition method for the character of man's face and ears, including the technique processes: get the images of the front and the side face, pick up the parameters of face structure character and the auricle size character through the front face image, and the parameters of the auricle shape and structure characters through the side face image, and combine the character attestation. This invention has the strong points of convenient method of getting image, the low cost of equipment, the stable character and the accurate attestation.

The invention discloses a footprint pressure distribution characteristic representation method which comprises the steps of (S1) using a footprint sampling instrument to collect footprint pressure image information, and preprocessing a footprint pressure image, (S2) extracting footprint front foot and back foot pressure distribution characteristics and constructing a footprint pressure distribution characteristic map, (S3) projecting front foot and back foot pressure distribution characteristic maps along direction, finding a direction with a maximum projection value (pressure value) as a main direction, and carrying out pressure distribution characteristic map main direction standardization, and (S4) carrying out characteristic selection on the obtained footprint pressure distribution characteristics. According to the method, the extracted footprint pressure distribution characteristics can represent the distribution characteristics of foot pressure when a human body is walking, the distribution between a footprint of barefoot and a footprint with socks is not needed, and the applicability is wide.

The invention discloses a low-cost barrier detection and tracking method for unmanned driving. The method comprises the following steps: acquiring environment point clouds, and overlapping the point clouds to obtain an environment point cloud set PC_0 of a current frame; removing points of which the height differences are larger than a threshold in a vehicle reference frame to obtain a point cloudset PC_1 consisting of non-ground points; filtering the length and width of the PC_1, keeping the environment point clouds within a fixed distance range, and converting the cloud points into a single-channel or three-channel feature image; inputting the obtained feature image into a built detection model to obtain an information list of a barrier target; building trackers for barrier targets of different types with different methods or parameters in combination with a barrier target information list of a previous frame according to the obtained barrier target information list, tracking each target in the current frame, and meanwhile calculating and updating motion information of the barrier targets; and repeating the steps 1 to 5 till the end of unmanned driving. The method has the advantages of low cost, high detection accuracy and high running speed.

The present invention relates to a binder resin for a toner which is used in electrophotography and the like. An objective of the present invention is to obtain a binder resin for a toner containing a crystalline resin which satisfies both excellent low temperature fixing property and excellent offset resistance, a method for producing the binder resin, and a toner using the binder resin. The objective can be achieved by using a binder resin for a toner that is produced by subjecting an amorphous resin and a crystalline resin to melting, kneading and reaction, and is characterized in that it includes a network structure which includes a crystalline resin.

The present invention provides a sea surface ship intelligent tracking system and method based on machine vision. The system comprises an image collection module, a Mean-shift tracking module, a Kalman filter module, a multi-stage parallel detection module and online learning module. The sea surface ship intelligent tracking system and method based on the machine vision employs a Haar+AdaBoost combination algorithm to detect the sea surface ships, and the detection result is taken as the starting frame of the tracker so as to initialize the tracker and the detector and replace the manual delineation of a target area. The Kalman filter is configured for the detection and the tracking of the ship so as to shorten the detection range, reduce the calculation amount and improve the timeliness. Based on the Mean-shift tracking module, the stable feature, anti-shielding, simple calculation and insensitivity of target deformation, rotation and background movement; the multi-stage parallel detection module is employed to combine the set classifier based on the random forest, the Haar classifier and the nearest neighbor classifier so as to solve the difficulty that the ship tracking is failed caused by shielding and deformation and improve the tracking timeliness, the accuracy and the robustness.

The invention discloses a method for estimating the physical dimension of a conical target of a broadband composite bistatic radar, and the method comprises seven steps: 1, building a composite bistatic radar observation model; 2, obtaining a monostatic / bistatic radar one-dimensional ranging image; 3, extracting the position of the scattering center of a monostatic / bistatic radar; 4, solving an intermediate variable through employing a target symmetrical structure; 5, estimating a bistatic angle through employing tracking data; 6, carrying out the combined solving of the height of a cone and the bottom radius; 7, carrying out continuous observation and solving a mean value. The method can directly obtain the physical dimension of an object, is not affected by the attitude of the object, is stable in characteristics, is simple and practical, and facilitates the quick extraction of the features of a slightly moving target.

The invention belongs to the technical field of identity information identification, and discloses a steady-state visual evoked electroencephalogram identity recognition method and system based on frequency domain coding. The method is characterized in that differential brain waves generated by different testees under the stimulation of the same frequency are taken as a basis; the difference relation and the difference range of the electroencephalogram signals on the fundamental wave and harmonic wave amplitudes under the plurality of electrodes are utilized to extract the characteristics of the EEG signals, the EEG characteristic information of each tested person is coded, after coding is completed, when new EEG information codes are input again, the new EEG information codes are comparedwith input codes, whether the new EEG information codes are information source persons input in advance or not can be judged, so that the purpose of identity recognition is achieved. According to themethod, the characteristics extracted by frequency domain coding are stable, the calculation is simple, the characteristics are obvious, the recognition rate is relatively high, an identity recognition system constructed on the basis of steady-state vision-induced electroencephalogram signals is more confidential and not easy to copy and forge, and compared with the complex processing, the efficiency of a computer is obviously improved.

An individual identification attestation method based on the auricle geometrical parameter, and adopts the connection of the vidicon and the computer, including the following processes: through the head image decide the position of the pupil of eyes; confirm the coordination origin of the head image, and carry out angle amendment to the head image; assure the mouth position, the two side borders of face, the upper, the lower, the external and the appointed frontier nodes of the auricle; calculate the relative width and the height of the auricle to get the parameter of the ears' characteristics; and match the ears' characteristics. This invention has the strong points of convenient method of getting image, the low cost of equipment, the stable character and the accurate attestation.

The invention provides a video target detection method based on a depth feature pyramid and tracking loss, and solves the problems of video target detection stability and accuracy. According to the method, a multi-scale feature map method is adopted in the target detection process, and the bottom feature map is high in dimension, rich in detail information, low in high-level feature latitude and rich in semantic information, so that the spatial information of the video image can be better utilized, and the detection can adapt to multi-scale and multi-type targets. According to the method, a multi-scale candidate window generation mode is adopted, small targets are subjected to dense sampling, large targets are subjected to discrete sampling, and independent processing, small-scale target fine sampling, large-scale target coarse sampling and detection speed increase are carried out according to target scales in consideration of different precisions required when targets with different scales are detected.

A method for deshadowing a laminographic image comprises selecting a mathematical morphology structuring element larger than examination elements of a laminographic image to be inspected, performing a mathematical morphological operation on said image, and differentiating a background of said image from said examination elements to remove said background. The mathematical morphological operation may be a dilation or a closing for a positive image. For a negative image the mathematical morphological operation may be an erosion or an opening. Embodiments call for the use of a power of two structuring element, a bi-directional power of two structuring element, a terraced structuring element, or a sloped structuring element. The mathematical morphological operation may comprise at least one piecewise linear dilation or erosion by a structuring element of limited support. Thresholding may be used to provide a deshadowed binary image.

The invention is the recognition method for the character of man's face and ears, including the technique processes: get the images of the front and the side face, pick up the parameters of face structure character and the auricle size character through the front face image, and the parameters of the auricle shape and structure characters through the side face image, and combine the character attestation. This invention has the strong points of convenient method of getting image, the low cost of equipment, the stable character and the accurate attestation.

The invention relates to a permutation aliasing image blind separation method and a device. The method comprises the following steps: obtaining a training data set according to a noisy permutation aliasing image of at least two known permutation region positions; the convolutional neural network being constructed, and the acquired training data set being input to the convolutional neural network for training, and the trained convolutional neural network being obtained. The permutation aliasing images to be detected are preprocessed and input to the trained convolution neural network to obtainthe image feature map. The image feature map is optimized, and the optimized image feature map and the original permutation aliasing image to be detected are multiplied to get the separation effect map. The invention adopts the convolution neural network to automatically extract the features of the replacement aliasing images, the extracted features are stable, are not influenced by human factors,and the accuracy of image separation is improved. Moreover, the separation process is simplified and the separation speed is improved by changing the separation problem to the classification problem.

The invention provides a chemical heredity epilepsy persistent state disease animal model and a construction method and application thereof. The epilepsy persistent state disease animal model is a mouse brain kernel group (a hippocampus CA1 region and a thalamus anterior nucleus VA region of a model I); injecting a brain stereotaxic virus (a chemical genetic virus rAAV-CaMKIIa-hM3D (Gq)-mCherry-WPREs-pA) in an apricot kernel BLA region and a thalamus anterior nucleus VA region on the outer side of a substrate of the model II, and embedding an electrode array in a mouse hippocampus CA3 region;after the mouse is recovered for one week, a metabolite CNO of clozapine is injected into the abdominal cavity so that the CNO is combined with a virus expression receptor, neurons are activated to induce epileptic persistent state attack, and the epilepsy persistent state disease animal model is obtained through behavioral observation and in-vivo multichannel local field potential recording judgment. The epilepsy persistent state disease animal model constructed by the method is stable in seizure duration, high in success rate, low in death rate and good in repeatability, and has important significance in researching the origin and formation mechanism of epilepsy persistent state, and screening and mechanism of drug-resistant epilepsy persistent state drugs.

In a MOS transistor, a structure of trenches or fins arranged in parallel to a gate length direction is formed in a stepwise manner along a gate width direction to thereby reduce a step height of each step. Even if the MOS transistor includes a deep trench or a high fin in order to increase driving performance per unit area, a uniform impurity concentration in a channel region, a source diffusion layer, and a drain diffusion layer can be made by an ion implantation method. Accordingly, there can be obtained a stable characteristic that variation in the characteristic due to a surface on which the channel is formed does not appear, and a lateral MOS transistor with high driving performance having a reduced on-resistance per unit area can be provided.

A hydrogen generation device includes: a reforming unit for steam-reforming raw material containing at least carbon atoms and hydrogen atoms to generate a hydrogen-containing gas; a raw material supply unit for supplying a raw material to the reforming unit; a steam generation unit for supplying steam to the reforming unit; a steam generation unit temperature detection unit for detecting the temperature of the steam generation unit; a heating unit for supplying a combustion gas for successively heating the reforming unit and the steam generation unit by heat transfer; and a control unit. The control unit controls a raw material supply amount from the raw material supply unit and a water supply amount from a water supply unit and controls one of the air amount to the heating unit, a fuel amount to the heating unit, and the raw material amount to the reforming unit so that no water remains in the steam generation unit according to the detected temperature from the steam generation unit temperature detection unit. Thus, it is possible to operate the hydrogen generation device without leaving water in the steam generation unit and to provide a hydrogen generation device operation method having a high reliability and economy and a hydrogen generation device using the same as well as a fuel cell generation system using the hydrogen generation device.

The invention discloses a screen shooting picture detection method and device, electronic equipment and a storage medium. The method comprises the following steps: firstly, obtaining a target picturecorresponding to a target area from a to-be-detected picture; then, on one hand, edge features are extracted from the target picture; and on the other hand, predicting the probability that the targetpicture contains the moire features by using a pre-trained probability prediction model, and finally, comprehensively analyzing whether the target picture is a screen shot picture or not in combination with the probability that the target picture contains the moire features in the edge feature extraction result. The method has the advantages of being stable in feature, high in classification precision and high in robustness in application of a picture auditing scene, especially a real estate certificate picture auditing scene, and the auditing cost can be effectively saved.

An active liquid crystal diffraction element includes: a first transparent substrate; a second transparent substrate; a liquid crystal layer; a first electrode; a second electrode; a control member; and an alignment member, wherein the second electrode includes a plurality of small electrodes; the control member controls a magnitude of an electric voltage applied to each of the plurality of small electrodes; the alignment member confers a liquid crystal molecular alignment without any pre-tilt to the liquid crystal layer; and the liquid crystal molecular alignment is parallel to the surface of the first transparent substrate.

The invention discloses a method for extracting characteristics of an infrared spectrum and identifying samples. The method comprises the following steps: collecting infrared spectrum signals of training samples by using an infrared spectrometer; extracting different frequency components of the infrared spectrum signals by using wavelet decomposition, setting a peak ratio of a special frequency as a characteristic and training the characteristic by classifying to obtain a classifier; collecting infrared spectrum signals of samples to be detected to obtain characteristics of the samples to be detected by using the same set rule, feeding the characteristics of the samples to be detected into the classifier, calculating and analyzing and then indentifying the samples to be detected. The method for extracting characteristics of the infrared spectrum and identifying samples has the advantages that an infrared spectrum characteristic extracting method based on wavelet decomposition is provided; when used for identifying the samples, the method is high in accuracy, quick in speed and stable in characteristic.

The invention discloses a radar HRRP target recognition method based on spectrogram segmentation preprocessing and a convolutional neural network. The method comprises the steps of firstly, carrying out pretreatment to reduce the sensitivity in an HRRP sample; carrying out spectrogram transformation on the sample; enabling the next CNN layer to extract the information contained in the HRRP envelope while retaining the sequence correlation contained in the HRRP sample; then extracting the high-level features through time sequence correlation modeling by the bidirectional RNN; and finally carrying out target classification through the softmax function.

The present invention relates to a binder resin for a toner which is used in electrophotography and the like. An objective of the present invention is to obtain a binder resin for a toner containing a crystalline resin which satisfies both excellent low temperature fixing property and excellent offset resistance, a method for producing the binder resin, and a toner using the binder resin. The objective can be achieved by using a binder resin for a toner that is produced by subjecting an amorphous resin and a crystalline resin to melting, kneading and reaction, and is characterized in that it includes a network structure which includes a crystalline resin.

The invention discloses a sign language recognition system based on sensing equipment and a working method of the sign language recognition system. The working method comprises the following steps: collecting joint point data of hands of a user in real time; performing noise reduction on the joint point data; segmenting the articulation point data after noise reduction to complete gesture action segmentation; calculating gesture features used for gesture classification; gesture features corresponding to the gesture actions are classified, and the types of the gesture actions are determined; and matching the sign language corresponding to the gesture action category, and displaying a sign language result. On the basis of gesture action modeling and gesture classification, an online sign language recognition system is designed and developed, and a solution is provided for communication between healthy listeners and deaf-mutes.

The invention relates to a multi-target detection method for a blood sample, which comprises the following steps: 1) preparing a reagent for reaction, wherein the reagent comprises a coding magnetic microsphere series, and the surface of the coding magnetic microsphere is connected with a specific capture antibody; wherein the encoding magnetic microsphere series comprises magnetic microspheres with different particle sizes, the particle sizes of the magnetic microspheres can be distinguished by a signal of a flow analyzer, and the encoding magnetic microsphere series is used for identifying the types of the microspheres; 2) carrying out reaction and treatment on a blood sample to be detected and a reagent, and removing interference components to form a fluorescent labeled reaction compound based on microspheres; and 3) sending the reaction compound into a flow optical detection system for analysis, obtaining a characteristic signal, identifying the type of the microspheres, and simultaneously detecting multiple targets in the blood sample. Compared with the prior art, the method has the advantages that the coding microspheres have stable characteristics, a whole blood sample is used, multiple detection is realized, and the like.

The invention discloses a method of inducing hymphomycetes to spore. The method comprises the following steps: (1) circular filter paper is taken, more than four filter paper holes are formed by uniformly shearing the filter paper, the filter paper is flatly placed in culture dishes and is soaked by adding a proper amount of distilled water, and the filter paper is taken out after high temperature and high pressure sterilization; (2) an agar medium is prepared and is sterilized, then the agar medium is added into the sterilized culture dishes, the addition amount of the agar medium in each culture dish is 15-20ml, and agar culture plates are manufactured after the agar culture medium is solidified; (3) the sterilized filter paper is flatly paved on the surface of each agar culture plate; (4) a proper number of hymphomycetes strains which do not spore are selected and are inoculated on the improved agar culture plate at the center position of each filter paper hole; (5) culture dish covers cover and are sealed, and the culture dishes are cultured for 10-30 days at room temperature under natural light rays. The method is simple to operate, the culture conditions are mild, the effect of inducing the hymphomycetes to spore is obvious, the method is effective for most hymphomycetes, conidiums are uniformly distributed on the filter paper, and sporing phenotypes are convenient to observe.

The invention discloses a double-sided detection device and method suitable for various automobile door planks, and the device comprises a detection table, the upper end face of the detection table is fixedly provided with a detection frame, the detection frame is provided with a suspension type clamping mechanism for fixing the automobile door planks, one side of the suspension type clamping mechanism is provided with a plurality of color CCD cameras, and the color CCD cameras are connected with the detection frame. The plurality of color CCD cameras are rotationally mounted on the detection frame. The device and the detection method have the advantages that a single-side visual system can be adopted, the purpose of detection is achieved by turning over the two faces of the door plate through the mechanism, the detection precision is high, compatibility is good, and cost is low.

The invention relates to a method and device for blind separation of permuted and aliased images. The method comprises: obtaining a training data set according to at least two permuted and aliased images containing noises with known positions of permuted regions; constructing a convolutional neural network to obtain The training data set is input to the convolutional neural network for training to obtain the trained convolutional neural network; the permutation and aliasing image to be detected is preprocessed and input to the trained convolutional neural network to obtain the image feature map; The image feature map is optimized, and the optimized image feature map is dot-multiplied with the original permutation and aliasing image to be detected to obtain the separation effect map. The invention adopts the convolutional neural network to automatically extract features from the permuted and aliased images, and the extracted features are stable and not affected by human factors, which improves the accuracy of image separation; and solves the separation problem by turning it into a classification problem, simplifying the The separation process is improved and the separation speed is improved.