103results about How to "Avoid features" patented technology

The invention relates to a mobile phone casing defect detecting method based on depth learning. The method comprises the steps that (1) the image of a mobile phone casing to be detected is acquired and pre-processed; and (2) the pre-processed image is input into a pre-trained defect detection model for defect detecting to acquire the position of a defect on the mobile phone casing, and the confidence of the position as the defect is provided. The defect detection model is a depth network based on depth learning, and comprises a feature extraction network and a classifier and regression device network, wherein the feature extraction network and the classifier and regression device network are in successive cascade. The feature extraction network carries out feature extraction on the pre-processed image to acquire a feature image. The classifier and regression device network classifies and regresses the feature image to acquire the defect position and the confidence of the mobile phone casing. Compared with the prior art, the method provided by the invention has the advantages of high detection precision and accurate and reliable detection result.

The invention discloses an aggregate digital image assessment system which comprises a motion control module, a computer module, an image acquisition module and a power supply module, an image preprocessing module, an aggregate identity module, an aggregate analysis and evaluation module and a data storage module. The assessment method includes that: a computer sends an instruction to a motion controller for controlling the linear sliding of a laser scanner; a CCD camera takes a picture of aggregate on a scanning board and transforms the image to a digital image; the digital image is transformed to a grey-scale map, And an edge detecting operator is utilized for carrying out image enhancement and restoration to the transformed grey-scale map so as to transform the grey-scale map to a binary image; the aggregate in the binary image is projected and the image of the aggregate in the binary image is detected and separated; IPP image processing and analysis software is used for calculating the three-dimensional coordinate of the surface of the aggregate; quantitative assessment category is carried out to the characteristics of the aggregate to obtain the gradation, the shape, the corner angle and the texture of the aggregate; the computer stores experimental data as well as analysis and assessment results. The aggregate digital image assessment system of the invention can be used for analysis and assessment of the aggregate.

The invention discloses a geographical science domain named entity recognition method, which is used for recognizing geographical science core term entities and geographical location entities. The method mainly comprises three steps of (1) establishing a geographical science domain dictionary, and using a new word discovery algorithm to identify new words in the geographical science domain in an unsupervised way; (2) training and testing based on a conditional random field (CRF) model and a multichannel convolutional neural network (MCCNN) model; (3) carrying out error correcting and fusion on entities recognized by the models by using a rule-based method. According to the geographical science domain named entity recognition method, the new words of the domain are identified as the dictionary in an unsupervised way by using the new word discovery algorithm, so that the work distinguishing effect is improved. The semantic vectors of the words are learnt from large-scale unmarked data in an unsupervised way, and basic characteristics of the words are synthesized and are taken as the input characteristics of the MCCNN model, so that manual selection and construction of the characteristics are avoided. The predicting results of the two models are fused by means of a custom rule, so that the problem of error marking in a recognition process can be corrected.

The invention discloses an image super-resolution method based on SAE and sparse representation, and belongs to the field of image processing. The image super-resolution method mainly comprises an off-line training stage and a test refactoring stage, wherein in the off-line training stage, image characteristics extracted by an SAE (Sparse Auto Encoding) model are subjected to dictionary training, and a dictionary pair reflecting corresponding relations of high-resolution images and low-resolution images is established; in the test refactoring stage, low-resolution images inputted by a user are subjected to super-resolution reconstruction by the obtained dictionaries and a sparse representation method. Through the application of the image super-resolution method, unsupervised learning training is performed on original image sampling data by using the SAE model, so that the defects that manually designed operator characteristic extraction is time-consuming and strenuous and the extracted characteristics are single are avoided, meanwhile, image characteristics represented by SAE compression are directly used for training of the high-low-resolution dictionary pair, the dictionary training is facilitated, lost detail components in the images can be estimated by the sparse representation method, and higher-quality high-resolution images can be restored from the low-resolution images conveniently.

The invention discloses a three-dimensional convolutional neural network (3D CNN) based video classifying method and belongs to the technical field of video processing. According to the method, a video is sampled at equal intervals to obtain a plurality of video segments, a video database is amplified, three-dimensional video segments are directly input into a 3D CNN, and time domain and space domain characteristics of the video are extracted, so that the limitation of a conventional video classifying method in manually selecting video characteristics and video modeling modes is improved. A parallel distributed 3D CNN multi-classification model lowers the complexity in learning the 3D CNN and enables a classification system to realize distributed parallel computation more conveniently. Relatively high identification rate can be achieved with only fewer video segments based on a 3D CNN multi-classification system, and videos not belonging to any type can be classified into new type, so that the classification error of the new type is avoided.

The invention relates to the technical field of image processing, and proposes a brain disease classification system based on a self-attention mechanism, aiming at solving the technical problems of low classification accuracy caused by the complicated process of preprocessing, feature extraction and feature selection of magnetic resonance images required in the classification and diagnosis of brain diseases. For this purpose, the brain disease classification system based on the self-attention mechanism in the invention comprises the following steps: pre-processing the acquired human brain magnetic resonance images of the brain disease patients to obtain the gray matter density map of the human brain; using a pre-constructed brain disease classification model to classify the gray matter density map to obtain a brain disease category of the brain disease patient, wherein The pre-constructed brain disease classification model is a three-dimensional convolutional neural network model basedon the self-attention mechanism. The system shown in the embodiment of the invention can quickly and accurately classify the categories of brain diseases.

The invention discloses a recurrent neural network (RNN)-based social network message bust detection method and system, and relates to the technical field of popularity prediction of contents in social networks. The method comprises the following steps: acquiring history messages published and forwarded by a user in a social network, and preprocessing the history messages to obtain a history forwarding time sequence; carrying recurrent neural network training on the history messages and the history forwarding time sequence, and generating a prediction model; and acquiring messages published and forwarded by the user in real time, generating a forwarding time sequence according to the messages, inputting the forwarding time sequence into the prediction model to generate feature expressions, inputting the feature expressions into a fully-connected neural network to carry out classification, and outputting the classification result in a softmax manner so as to complete the social network message burst detection.

The invention relates to vehicle engine high-speed starting and stopping control method and system based on an active shutdown mode. The method comprises steps of: 200, making an engine work normally; 201, judging whether the current speed is zero, if so, executing 202 and if not, returning to 200; 202, judging the current gear, executing 203 if the gear is gear N, executing 204 if the gear is gear D and keeping idling if the gear is gear P; 203, judging the setting time of the gear N, executing 205 if the setting time is larger than a demarcate limit value, and otherwise, returning to 200; 204, judging if the last speed is larger than a certain demarcate limit value, if so, executing 205, and if not, returning to 200; and 205, judging the safety factor information of the current engine, executing 217 to shut down the engine if reaching a safe value, and returning to 200 if not reaching the safe value . The control system comprises a signal acquisition unit and a signal control unit, wherein the signal acquisition unit comprises an engine safety factor information acquisition module and a gear information acquisition module. Therefore, the invention can realize high-speed startingand stopping of the engine by a controller according to the working condition.

The invention relates to an electromagnetic-signal-time-inversion-based fault point positioning method for a transmission line. On the basis of an electromagnetic time inversion principle, a single observation point is set at any terminal in a power distribution network and a fault high-frequency transient signal within a limited period of time is measured and recorded; and time inverted sequenceprocessing is carried out on the measured signal, the processed signal is injected into a power distribution network simulation calculation model at an observation point, norm values at different guess positions in the power distribution network are calculated, and a fault position is determined by searching for a maximum peak value and a maximum energy root value. According to the method different form the traditional traveling wave determining method and impedance method, a power distribution network with a complex topological structure can be covered by one observation point without being affected by the fault impedance; and the excellent anti-noise performance is realized. The electromagnetic-signal-time-inversion-based fault point positioning method has the great practical value in fault positioning of the transmission line.

The invention relates to a landmark dynamic configuration method and device for a mobile robot searching unknown environments. A landmark is the positioning and navigation indication of a robot, no landmark arranged in advance exists in the unknown environments, and natural characteristics are hard to extract from the environments for serving as the landmark. The scheme of the method for solving the problems is as follows; a robot with a dynamic configuration landmark enters the unknown environments for searching, the landmark is independently and dynamically configured in an on-line way (realized by a vehicular putting device, seeing the abstract drawing) according to the positioning and drawing establishment requirements, and the configured landmark is used for finishing the positioning and environment drawing establishment of the robot. A dynamic configuration landmark positioning device realizes the positioning of the robot on the basis of an ultrasonic distance-measuring principle, realizes landmark identification on the basis of an exclusive identity method, and realizes data alternation on the basis of radio frequency communication, thereby having the advantages of high precision, small volume, low power consumption, and the like. The invention can be used for the fields of military affairs, antiterrorism, extreme environment operation, and the like. which requires intelligent equipment to carry out independent operation under the unknown environments.

The invention discloses a multistage video action detecting method. The multistage video action detecting method comprises the steps that rough action fragments fusing multiscale sampling and single-scale training are generated by an input video which is not clipped through a dichotomy and voting fusing strategy based on a deep residual network; the rough action fragments are subjected to action classification and action boundary joint distinguishing through a statistical fusion strategy based on a frame-level action recognition result, and primary action detecting fragments are obtained; in combination of IoU between the primary action detecting fragments, an improved non-maximum suppression algorithm is utilized for processing the primary action detecting fragments, and finally the action detecting result, namely, the action classification and beginning and ending time position of each video action detecting fragment, of the video which is not clipped is obtained. According to the method, the action classification accuracy and action locating precision can be improved.

The present invention relates to a method and a system for matching MR image feature points before and after the nonlinear deformation of a biological tissue. According to the technical scheme of the invention, a feature point automatic detection method based on a depth-cascaded convolutional neural network is provided. According to the method, firstly, a general region of feature points is obtained through the first layer of the depth convolutional network. Secondly, the position of a target feature point is approximated step by step in the second and third layers of the cascade convolutional network, so that the detection rate of feature points is further improved. The method aims to solve the problem in the prior art that the feature point distinguishing ability is reduced due to the image nonlinear deformation of existing feature point descriptors. In this way, a Riemannian manifold is combined with the kernel method to construct a nonlinear deformation feature point descriptor for robustness. The three-dimensional feature points of a magnetic resonance image are mapped into a four-dimensional Riemannian manifold space. Meanwhile, the feature points are further mapped into a higher-dimensional Hilbert space based on the kernel method, so that a richer description of data distribution is obtained. Meanwhile, a real geometric distance between feature points is obtained, so that the feature points are matched.

The invention discloses a method and device for generating a special effect video, belonging to the technical field of computers. The method includes the following steps: receiving a special effect adding instruction, and recording special effect information corresponding to the special effect adding instruction; during video recording, when a recording completion instruction is received, ending the video recording, and displaying a video editing page; and when a special effect editing completion instruction triggered by the video editing page is received, performing special effect editing onthe recorded target video according to the special effect information to obtain a special effect video corresponding to the target video. By adopting the scheme of the invention, the occurrence of thelow-speed operation of a terminal can be reduced.

The invention provides a method and device for detecting and identifying stored grain insects. The method comprises the steps of inputting an original stored grain insect image into an artificial intelligence analysis model to perform location positioning and category judgment on insects, wherein the artificial intelligence analysis model is a convolutional neural network which is trained and verified according to a data set formed by the stored grain insect image; and labeling the insect location and category on the original stored grain insect image according to a location positioning and category judgment result. According to the invention, a model capable of correctly positioning and identifying various stored grain insects is trained according to a deep leaning based target detectionalgorithm and collected stored grain insect data, thereby avoiding the drawbacks of poor robustness and generalization of manual design features, and improving the accuracy and efficiency of detectionand identification.

The invention discloses a piezoelectric elastic wave deicing method. A signal generator generates two paths of alternating voltage which have phi / 2 phase difference and are same in frequency, the two paths of alternating voltage are respectively amplified by two power amplifiers, two piezoelectric units which are attached on the inner surface of an aircraft skin are used for exciting two piezoelectric vibrators in different vibration type on the skin, i.e. the vibrations of the first vibration type of piezoelectric vibrators and the second vibration type of piezoelectric vibrators, under the excitation of the same driving frequency, the first vibration type of piezoelectric vibrators and the second vibration type of piezoelectric vibrators generate two vibration modes, i.e. longitudinal vibration or bending vibration, the two vibration modes are stacked on elastic waves formed on the aircraft skin, the elastic waves are similar to traveling waves, the shearing force is generated on the attaching interface of an ice layer, the ice layer is forced to strip from the skin under the action of the shearing force, and then the aircraft surface is deiced. The piezoelectric elastic wave deicing method has the advantages that the energy consumption is low, the power consumption is low, the piezoelectric elastic wave deicing method is compatible with an aircraft avionics system and does not require single power supplying, and the continuous operation is realized.

The present invention relates to a radio signal identification method based on an end-to-end convolutional neural network. The method is characterized in that: an original I / Q sampling data of an observation window is subjected to execution of preprocessing and identification through a convolutional neural network in order. The preprocessing step is that: the original I / Q sampling data of the observation window is taken as input, and a frequency spectrum waterfall plot is output after discrete Fourier transform and data format alignment processing; the step of identification through the convolutional neural network is that: the frequency spectrum waterfall plot obtained by preprocessing is taken as input, and a one-dimensional boolean vector configured to show whether all the signals to beidentified are existed or not is output after the input passes through a CNN feature extraction layer, an MLP feature mapping layer and a BR multi-tag classification layer. Compared with the mode offeature extraction and classification identification, the radio signal identification method employs the end-to-end technical solution thinking to avoid complex and low-efficient feature engineering,improve the signal identification accuracy, robustness and intelligence level, and has important meaning of radio monitoring of important areas and important activity scenes.

The invention relates to a high-speed air maneuvering target detection method based on STAP (Spacetime Adaptive Processing), belonging to the technical field of maneuvering target detection. The detection method comprises the following steps of: (1) estimating a clutter covariance matrix of a unit to be detected according to data of a reference unit; (2) multiplying an inverse matrix of the clutter covariance matrix with received data to realize clutter suppression; (3) performing Keystone conversion correction target distance walking on data which are subjected to clutter suppression; (4) performing airspace beam forming on data which are subjected to Keystone conversion correction target distance walking; (5) performing Wignet-Hough conversion on the obtained data to estimate a target acceleration; (6) performing Doppler walking compensation on the acceleration obtained by estimating the data which are subjected to Keystone conversion correction target distance walking obtained in the step (3) according to the step (5); and (7) performing spacetime two-dimensional beam forming on the obtained data to realize target accumulative detection. The method has the advantages of easiness, accurate detection, suitability for severe distance walking of the target, movable target detection during Doppler ambiguity and Doppler walking, and the like.

The invention discloses a method for estimating air particulate pollution degree based on shallow convolution neural network. The basic steps of the method include: 1. Constructing a shallow convolution neural network (PMIE) model with a layer enhancement function; 2. Combining the output of PMIE model with four kinds of weather eigenvalues to construct regression model; 3. Trainning PMIE model and regression model; 4. The PM2.5 index of the test set image is estimated by using the trained PMIE model and regression model. The invention provides a shallow convolution neural network model with layer enhancement function, Combining the output results with four kinds of weather features to estimate the degree of air particulate pollution in the image, the problem caused by feature extraction and feature optimization is effectively avoided, and the specific PM2.5 index value is obtained, which improves the training convergence speed and algorithm robustness, and has better performance.

The invention discloses a method for extracting blood pressure data from a pulse wave signal and a device thereof. The method comprises: model training and blood pressure data extraction; wherein themodel training comprises: acquiring the pulse wave and corresponding blood pressure data, and performing preprocessing; performing complete waveform detection and separation of the pre-processed pulsewave; performing frequency domain transformation on each complete pulse wave waveform, and extracting the frequency domain features; taking the frequency domain feature as an input, taking the corresponding blood pressure value as an output, combining with the neural network for predictive model training, and obtaining a blood pressure data prediction model. The blood pressure data extraction includes the following steps: inputting the frequency domain feature to-be-predicted into a blood pressure data prediction model, and outputting a blood pressure value. The pulse wave is detected and separated, the frequency domain is transformed to extract features, which can effectively avoid the problem of difficult feature detection in a time domain feature extraction method, provides more accurate samples for subsequent model training with neural network, and improves the prediction accuracy of blood pressure measurement and reduces the prediction error.

The invention discloses a digital display instrument reading identification method based on a convolutional neural network in the technical field of mode identification and artificial intelligence. The digital display instrument reading identification method comprises the steps of obtaining data, processing the data, constructing a depth network model, reading instrument reading and the like. According to the digital display instrument reading identification method based on the convolutional neural network disclosed by the invention, a high-precision automatic digital display instrument reading identification method is realized through a learning and training process based on a big-data instrument image; and the method has the characteristics of being high in identification accuracy rate and real-time performance, better in practical value and the like.

The invention relates to the field of hole decorating plate machining, in particular to a hole decorating plate machining method based on gray levels and a machining system applying the hole decorating plate machining method based on the gray levels. The hole decorating plate machining method based on the gray levels comprises following steps of (1) obtaining a custom image; (2) conducting a gray processing treatment on the image; (3) segmenting the image into areas; (4) defining the parameters of holes of area linkages; (5) transforming machining technology data; and (6) generating numerical control machining programs. According to the machining method and the machining system, treatment work of the preparation link before production is integrated by using an image treatment technology; the image characteristics are refined and the imaging effect is strengthened by treating the image combining a ''gray processing'' method and mapping ''gray level values'' with punching characteristics, so that automatic ''data mining'' of a computer can be realized, the problems such as rough imaging can be avoided and the imaging quality of the products can be improved; the consumption of labor force can be reduced, the cost can be lowered and the production and use of the products can be facilitated conveniently by the function of integrating the automatically generated numerical control machining programs.

The invention provides a stroke segmentation method based on a mixing feature for intelligent free-hand input. The method comprises the steps that sampling points of effective input strokes are extracted; speed segmentation points of the effective input strokes are obtained by the adoption of a stroke segmentation method based on a speed feature, and polygonal approximation is performed on the effective input strokes to obtain a salient point sequence; primary feature mapping is performed on the salient point sequence and a speed segmentation point sequence to obtain a speed salient point sequence; the effective input strokes are segmented to obtain a geometric segmentation point sequence; secondary mapping is performed on the speed salient point sequence and the geometric segmentation point sequence to obtain an early-stage segmentation point sequence; early-stage segmentation points are judged to obtain late-stage segmentation points. According to the stroke segmentation method, the speed feature and the geometric feature are combined to form the new mixing feature after mapping; firstly, early-stage segmentation is performed, afterwards, the single line element recognition is performed, and finally late-stage segmentation is performed. In this way, the defects of simple geometric feature segmentation and the defects of simple speed feature segmentation are effectively avoided.

The invention discloses a temperature sensitive polymer material with biodegradability and biocompatibility. A preparation method of the material comprises the following steps of: firstly synthesizing monomers with specific structures by using the basic raw materials including lysine, tartaric acid, and malic acid, and carrying out polymerization reaction on the monomers with different ingredients to prepare the heterochain type polymer material. The polymer not only has good temperature sensitivity, but also has good biodegradability and biocompatibility. The preparation method disclosed by the invention is simple and easy to implement and has good industrial prospect.

The invention relates to the technical field of data processing, in particular to a gear broken tooth detection method based on Hough circle detection. According to the method, a pixel code of each line is obtained through line-by-line analysis of a gear binary image; and obtaining a second pixel value sequence corresponding to the pixel code which is the same as the standard code. According to the gear center hole pixel value position in the second pixel value sequence, hole edge pixel points are obtained, and then the circle center is obtained. Performing Hough circle detection on the circle center to obtain a plurality of initial reference circles, and screening out reference circles according to the number of gear information on the edges of the initial reference circles. According to the number of the pixel points of the gear information and the background information on the edge of the reference circle, whether tooth breakage occurs is judged. According to the method, the edge acquisition process is simplified in pattern recognition by combining related electronic equipment, and the detection of the position and degree of the broken tooth is quickly and effectively realized through the analysis of the pixel value on the reference circle.

Provided is a method for producing liquefied natural gas by multi-component refrigerant two-stage compression. The method includes two multi-component refrigerant compression cycles, the refrigerant in a first cycle is called a pre-cooling refrigerant, the refrigerant in a second cycle is called a cryogenic refrigerant; the pre-cooling refrigerant returns to an ice chest after passing through a pre-cooling compressor, a cooler, a buffer tank, the ice chest, a throttle valve and a gas-liquid separation tank, and then returns to the pre-cooling compressor; the cryogenic refrigerant returns to the ice chest after passing through a cryogenic compressor, the cooler, the ice chest, the throttle valve and the gas-liquid separation tank, and then returns to the cryogenic compressor; and natural gas is cooled after heat transfer of the ice chest and the refrigerants and finally becomes the liquefied natural gas. The natural gas is liquefied by two independent multi-component refrigerant compression cycles, and composition and content of the two multi-component refrigerants are different, so that the method is wide in application range to raw material natural gas, high in operation flexibility and high in liquefying efficiency. Besides, the method is simple in liquefying process, low in investment cost, and is applicable to liquefied natural gas factories in various scales.

The invention discloses an attention mechanism-embedded iterative aggregation neural network high-resolution remote sensing scene classification method. The method comprises the steps of firstly performing convolution, channel attention screening and fusion on a high-resolution remote sensing image by adopting an iterative aggregation module with attention to obtain bottom-layer features of the image; secondly, carrying out convolution on the obtained bottom-layer features, and then sending the bottom-layer features to a next iterative aggregation module with attention to extract middle-layerfeatures of the image; carrying out convolution on the obtained middle-layer features, and then sending the middle-layer features to the last iterative aggregation module with attention to extract high-layer features of the image; finally, classifying the remote sensing scene images by the feature map through a pooling layer and a full connection layer. According to the method disclosed in the invention, the remote sensing image is subjected to feature extraction and fusion by using the iterative aggregation module with attention, the extracted features cover relatively strong semantic information, and meanwhile, a structure with an attention mechanism is embedded in the module, so that channels with useful information can be effectively screened out for fusion, and the recognition capability and classification performance of a classifier are improved.

The present disclosure relates to the field of computers, and provides a data processing method and device, a computer readable storage medium and an electronic device. The data processing method includes: acquiring a plurality of user location information of a user within a preset time period to determine a first trajectory of the user; acquiring the same biological characteristic information acquired by image acquisition devices at different locations during the preset time period so as to determine a second trajectory of the biological characteristic information; and using the biological characteristic information as the biological characteristic information of the user if the first trajectory matches the second trajectory. The data processing method can more accurately match a user with his or her biological characteristics, and improves the accuracy of the characteristic library.

The invention provides an industrial internet alarm log association analysis method and system based on a graph method. The method comprises the following steps: step 1, obtaining a network attack alarm log; step 2, analyzing the obtained network attack alarm log to obtain a characteristic quantity corresponding to each network attack alarm; 3, creating a network security event graph according to the obtained characteristic quantity, and dividing the network security event graph into a plurality of alarm clusters; 4, extracting statistical characteristics and topological structure characteristics corresponding to each alarm cluster; 5, analyzing and identifying each alarm cluster according to the obtained statistical characteristics and topological structure characteristics; according to the method, the influence of multi-source and heterogeneous characteristics of alarm log data is avoided, the processing efficiency of common security event alarms and the recognition capability of high-risk events can be effectively improved, and the perception capability of the overall security situation of the industrial internet can be provided.

The invention relates to a license plate Chinese character recognition method. The license plate Chinese character recognition method includes steps of manually constructing a standard training sample gray image; preprocessing the training sample gray image to generate a preprocessed gray image; performing band-pass filtering for the preprocessed gray image by the aid of P Gaussian band-pass filters with interconnected pass bands to obtain P filtered gray images; performing dimensionality reduction for the P filtered gray images by a linear or nonlinear process to form images with M-dimensional training sample characteristic vectors; and performing license plate Chinese character recognition for the images with the M-dimensional training sample characteristic vectors by the aid of an RBF (radial basis function) neural network after the images are subjected to dimensionality reduction, and outputting Chinese characters of license plates. The method has the advantages that the RBF neural network is adopted, network parameters are solved by an unsupervised clustering and least square process for solving linear equations, repeated iteration is avoided in a solving procedure, the time efficiency is high, algorithm convergence is good, and the method is high in generalization ability.

The invention belongs to the technical field of communication, and specifically relates to a modulation recognition method based on neural network ensemble. The modulation recognition method based onneural network ensemble utilizes a convolutional neural network to automatically extract the integrated abstract features, avoids the design and selected signal characteristics of a traditional method, and in fact, can obtain different classifiers in the mode of changing a training set so as to adapt to most of modulation modes. In addition, the modulation recognition method based on neural network ensemble uses the ensemble strategy to enhance the recognition performance at a low signal to noise ratio.