60results about How to "Preserve features" patented technology

The invention discloses a method for reconstructing a three-dimensional curve face through point cloud. The method comprises the steps of inputting point cloud data P possibly carrying noise and abnormal values and the number m of peaks of the curve face needed to be reconstructed; initializing a dictionary matrix V and a connection matrix B; updating the dictionary matrix V and the connection matrix B in an iteration mode till convergence; outputting a reconstructed triangular net to complete reconstruction of the three-dimensional curve face. By means of the method, the triangular net is directly reconstructed through the point cloud, the noise input into the point cloud can be removed well, non-uniform sampling is processed in a robust mode, and the characteristics in the point cloud can be recovered well. Compared with a conceal method, the triangular net is directly reconstructed through the point cloud so as to avoid accumulated errors caused by multi-step optimization in the conceal method. Compared with a combination method, the reconstruction errors serve as a target function, so that the reconstruction errors of the curve face reconstructed through the method are usually smaller than those reconstructed by the existing combination method.

The invention discloses an operating system vulnerability detection method and system based on code similarity analysis and a medium. The method comprises the steps: locating a vulnerability code segment, and forming an operating system vulnerability code library; screening an operating system source code library; for the vulnerability code segment, respectively generating a code attribute graph set by an operating system source code library; and respectively extracting features from the vulnerability function in the vulnerability code segment and the code attribute graph set of each functionto be tested in the operating system source code library, calculating similarity, detecting whether reuse of the vulnerability code exists or not, and outputting a result. The method aims at solving the problems that an existing vulnerability code reuse detection method is insufficient in detection capacity when oriented to large-scale software systems such as an operating system. By adopting themethods of screening the operating system source code library, progressively screening the multi-vulnerability characteristics and the like, the process of the existing detection method is improved and optimized, and good accuracy is considered while high operating efficiency is ensured.

The invention discloses a remote sensing image fusion method. The method comprises the following steps of: constructing three-layer pyramid structures for panchromatic images and multi-spectral imagesrespectively, and sampling down panchromatic images and multi-spectral images of the three-layer pyramid structures respectively to obtain the re-extracted panchromatic images and re-extracted multi-spectral images corresponding to the three-layer pyramid structures; according to the spectral energy relationship between the re-extracted panchromatic images and the re-extracted multispectral images, the preliminary fusion coefficients of the remote sensing images corresponding to the re-extracted panchromatic images and the re-extracted multispectral images with the same pyramid structure arecalculated. According to the restored multi-spectral image, the object type of each pixel in the restored panchromatic image is determined. The initial fusion coefficients of pyramid structures are adjusted according to the geographic features, and the final fusion coefficients are obtained. The panchromatic image and multi-spectral image are fused according to the final fusion coefficients. The remote sensing image fusion method of the invention can improve color fidelity and clarity and enhance detail information of the fusion image.

The invention relates to a video gesture classification method. The method comprises the following steps: the visible light video and the depth video of the same object set are normalized; the saliency of each frame in the visible light video is calculated to obtain the saliency video; the three-dimensional convolution neural network is respectively trained by using normalized visible light video,depth video and salient video, wherein the three-dimensional convolution neural network comprises an input layer, 18 convolution layers, a global average pooling layer and a full connection layer arranged in turn, the features of the global average pooling layer of the spatial dimensions of the visible light video, depth video and salient video are fused to obtain the fused features, and the video gestures are classified according to the obtained fused features. The invention aims at the problem of large-scale gesture recognition based on video, and the category data and the video data amountthereof are far larger than those of the prior invention. The invention can carry out real-time identification on the video, and has more credibility and practicability.

The invention discloses a bank note one-dimensional signal detection method and device. According to the bank note one-dimensional signal detection method and device, bank note one-dimensional signal data are converted into a two-dimensional gray value; the two-dimensional gray value is preprocessed; and the two-dimensional gray value is detected through using a two-dimensional data processing method, so that whether an adhesive tape is bonded on a bank note can be detected, the authenticity and face value of the bank note can be detected. According to an existing one-dimensional signal detection, abnormal samples are difficult to be distinguished, and normal samples are falsely denied, and bank notes bonded with adhesive taps are difficult to be identified accurately, while with the bank note one-dimensional signal detection method and device of the invention adopted, the problems in the prior art can be solved.

The invention discloses a small-sample face recognition method combining sparse representation and a convolutional neural network, and the method comprises the following steps: firstly carrying out the preprocessing of a face image, carrying out the face alignment and five-sense-organ positioning according to face key points, and cutting the face image into four local regions; extacting local features and overall features with higher discrimination by using a convolutional neural network, and constructing a block feature dictionary in combination with a sparse representation algorithm, so as to achieve the effect of sample enhancement; adding sparse representation constraints and cosine similarity to redefine a loss function of the convolutional neural network so as to reduce the intra-class distance between the features and expand the inter-class distance; and finally, carrying out face recognition by adopting enhanced sparse representation classification. The method is high in recognition performance, and has certain robustness for shielding changes in a small sample problem.

The invention discloses a thermal power plant time sequence variable abnormity monitoring method and system. The method comprises the steps of collecting time sequence data of thermal power plant analog signal monitoring variables; taking the whole time sequence data as an initial optimal PLR segment, calculating a decision coefficient of the optimal PLR segment, and taking the whole time sequencein the historical sample data as a first-layer original data segment; segmenting the original data segment to form two new data segments, and respectively calculating decision coefficients of the twonew data segments; judging whether the segments are allowed or not, and if yes, continuing to segment the segments; and judging whether the segmentation is finished or not to obtain the optimal PLR segmentation number. According to the method, the optimal segment number of the PLR is determined through a coefficient determination method, and the problems that an L method highly depends on a fitting error and left and right points in a time sequence are solved.

The invention discloses a six-degree-of-freedom hydraulic mechanical arm, and belongs to the technical field of mechanical equipment. The six-degree-of-freedom hydraulic mechanical arm comprises a base rotary joint, a big arm, a first telescopic joint, a first mechanical arm, a second telescopic joint, a second mechanical arm, a horizontal swing joint, a rolling swing joint, a rolling swing support, a pitching joint and two bearing assemblies; one end of the big arm is fixed to the top surface of the base rotary joint, and the other end of the big arm is rotationally connected with one end ofthe first mechanical arm through the bearing assemblies; the other end of the first mechanical arm is rotationally connected with one end of the second mechanical arm, and the other end of the secondmechanical arm is fixedly connected with the horizontal swing joint; the rolling swing support is rotationally connected with the horizontal swing joint, the rolling swing joint is arranged on the inner side of the rolling swing support, and the pitching joint is rotationally connected with the rolling swing support; the two ends of the first telescopic joint are rotationally connected with the big arm and the first mechanical arm correspondingly, and the two ends of a second telescopic joint are rotationally connected with the first mechanical arm and the second mechanical arm correspondingly. According to the six-degree-of-freedom hydraulic mechanical arm, hydraulic driving is adopted, so that the working angle can be accurately adjusted.

The invention provides an advertisement image detection method and device. The advertisement image detection method comprises the steps of extracting images at reference positions in advertisement video stream and at least one image before and after the reference positions; taking pixels with high identification rates in the selected images as identification points for matching images; segmenting the extracted images in a mode of dividing the images at the reference positions and the at least one image before and after the reference positions of the extracted images into segments, thereby forming image segments; dividing the image segments into multiple parts, thereby forming image groups; carrying out parallel similarity matching on the identification points of the reference position images in each image segment of the image groups and the identification points of the at least one image before and after the reference positions, thereby obtaining an advertisement segmentation time vector of each image group; and obtaining start time and end time of each piece of advertisement in the advertisement video stream according to the advertisement segmentation time vector of each image group.

The invention discloses an SLAM method based on fusion of a monocular vision feature method and a direct method, and the method comprises the steps of extracting DSO features from a gray level image,extracting ORB features from the DSO features, and carrying out the feature pair matching of the DSO features and the ORB features; inserting the key frame into the local map, updating the existing map points, removing redundant map points, extracting new map points from the key frame, and updating the map points; projecting the existing map points to the key frame, constructing local bundle adjustment, and adjusting the pose of the camera; removing redundant key frames, inquiring the database, carrying out Sim < 3 > similarity calculation, detecting whether a closed loop exists or not, if so,carrying out closed loop fusion, optimizing an essential diagram, and ending the process; if not, entering the next step; and constructing a global bundle adjustment, adjusting the pose of the camerato minimize the overall error of the SLAM system, and updating the map.

The invention relates to a gel-like oviductus ranae health-care food. The health-care food comprises oviductus ranae and jelly composited with the oviductus ranae. The jelly is prepared from colla corii asini or tortoise-plastron glue or gelatin or mixture of the colla corii asini, the tortoise-plastron glue and the gelatin. The health-care food has the functions of delaying senility, resisting fatigue, tonifying the kidney to arrest spontaneous emission, moistening the lung and nourishing yin (Chinese character), and regulating and improving coronary heart diseases, and has good taste and appearance shape.

The invention relates to a fire image edge extraction processing method, a terminal and a system, and belongs to the technical field of fire flame extraction. The method comprises the steps of obtaining a fire image of a target object; performing pixel screening on the fire image by using pixel RGB and HSV information to obtain a first fire binary image of the target object, converting the fire image into a grayscale image, and obtaining a second fire binary image of the target object by using an Otsu algorithm; performing AND processing on the first fire binary image and the second fire binary image to obtain a third fire binary image of the target object; and performing edge extraction on the third fire binary image to obtain a first target flame edge extraction image of the target object. Compared with a classic canny algorithm and a local interference removal improved canny algorithm, the fire image edge extraction effect is better, and therefore the accuracy of fire flame extraction can be improved.

The invention relates to a face image restoration method based on a face pattern, and the method comprises the following steps: constructing a face pattern-based generative adversarial network which comprises a face pattern-based generative network and a PatchGAN discriminator network; wherein the face pattern-based generative network comprises a trunk repair sub-network formed by an encoder and a decoder, and a face pattern extraction sub-network capable of extracting a local face region pattern and a global face pattern of each face region according to a face analysis graph; wherein all the local face area styles and the global face styles form a style matrix, and affine parameters of all the face areas are generated; and hopping connections exist between corresponding network layers in the encoder and the decoder, and a face area restoration module is embedded in each hopping connection. The problems of visual artifacts, fuzzy facial structure details and inconsistent facial semantics in the repaired face image in the prior art can be effectively solved, and a more accurate repairing effect is obtained.

The invention relates to a photovoltaic power generation power day-ahead prediction method based on image feature extraction. The method comprises the steps of firstly collecting historical data of each influence factor of photovoltaic power generation, wherein the historical data comprises temperature, clearness index, PM index, power generation power, humidity, wind speed and the like; constructing a column matrix taking each influence factor as a row, and taking matrix elements as picture pixel points to obtain a plurality of two-dimensional images; carrying out normalization processing onthe image according to rows, setting convolution kernel dimensions and number, moving strides and filling dimensions, dividing photovoltaic power generation power into a plurality of gears according to accuracy, and carrying out category marking on the gears; and extracting image features based on a convolutional neural network principle, and establishing a prediction model to realize photovoltaicpower generation power day-ahead prediction. According to the photovoltaic power generation power prediction method provided by the invention, the prediction speed is improved while the prediction precision is improved, the influence of photovoltaic power generation power fluctuation on a power grid is reduced, and the photovoltaic power generation power prediction method has important significance for popularization and application of photovoltaic power generation.

The invention discloses a method for extracting the fault signal feature of a mechanical device. The method comprises steps of acquiring, by a sensor, the state signals of respective components of themechanical device; performing empirical wavelet transform on the acquired state signals to obtain the modal components of signal decomposition; calculating the correlation coefficients between respective modal components and an original signal, and selecting a plurality of modal components having the highest correlation with the original signal as the optimal modal components; constructing a Hankel matrix for the optimal modal components, and then performing singular value decomposition to obtain the feature values of the optimal modal components; subjecting the feature values of the optimalmodal components to dimensionality reduction, and combining the feature values subjected to the dimensionality reduction into a fault feature vector. The method can effectively remove the noise in theoriginal signal, retains the useful features in the original signal, and provides a necessary basis for the fault diagnosis of the mechanical device.

The invention discloses a power load clustering method and system. The power load clustering method comprises the steps: obtaining a load time sequence of each user, inserting a phase shift into the load time sequence of each user, and obtaining an embedded sequence; generating mapping function parameters for the embedded sequence by adopting a linear neural regression method; and clustering the users by utilizing the similarity of the mapping function parameters of the users to obtain a typical load curve. According to the power load clustering method, load time series data with asynchronousacquisition time, unequal time lengths and noise can be processed.

The invention belongs to the technical field of lithology identification, and provides an unmanned aerial vehicle-mounted XRF and image fusion lithology identification system and method. The lithology identification system comprises an information acquisition system, an information fusion system and a lithology identification system. The information acquisition system, the information fusion system and the lithology identification system are all carried on the unmanned aerial vehicle; the information acquisition system comprises an XRF detection system and an image acquisition system which are used for detecting element information and image information of rocks in a set area and transmitting the element information and the image information to the information fusion system; the information fusion system is used for expanding the element information into three-dimensional element information and fusing the three-dimensional element information with three-dimensional image information to obtain fusion information; the deep features of the image information and the shallow features of the fusion information are fused step by step to obtain a fusion feature map; and the lithology identification system is used for identifying the lithology of the rock in the set area based on the fused feature map.

The invention discloses a low dimensional successive projection pursuit cluster (LDSPPC) model comprehensive evaluation method, device and application. The method comprises a step of carrying out normalized preprocessing on sample data of multiple candidate objects, and constructing a one-dimensional projection pursuit cluster model with 2 to 4 mutually orthogonal projection vectors for the candidate objects, and a step of carrying out projection pursuit cluster model vector synthesis of all dimensions of the multiple candidate objects to form a comprehensive projection pursuit cluster model, and obtaining an evaluation index importance ranking list and a candidate object quality ranking list. The swarm search intelligent algorithm of the invention has the characteristics of a fast convergence speed, convergence to a globally optimal solution and high reliability, the vector synthesis of multiple successive projection pursuit vectors is carried out, the quality of the candidate objects can be quickly evaluated, and the accuracy of candidate object quality evaluation is improved.

The invention relates to a peak noise suppression method of variable-window weighted seismic data and aims at solving various problems that peak noise has characteristics of high energy, occurrence time randomness and wide frequency band, a noise source is complicated and variable to cause that a plurality of peak noises have different continuous time and frequency band widths, the peak noise cannot be favorably suppressed by a fixed-window peak noise suppression method and the applicability is poor. Through experiment, the method has a favorable suppression effect on seismic data disturbed bystrong peak noise with different continuous time, the peak noise is effectively suppressed, and the characteristic of a signal can be maintained according to a detected peak noise changing window length autonomous estimation weight. Compared with an existing mid-value substitution method, the peak noise suppression method has the advantages that the resolution of identifying the peak noise is higher, and the signal energy is fully maintained; and meanwhile, the peak noise suppression method is applicable to different environmental situations, the characteristic of the peak noise is analyzed in real time, the peak noise is suppressed, and the quality of acquired data is effectively ensured.

The invention relates to the technical field of micro-seismic monitoring, and concretely relates to a mining micro-seismic signal filtering method. The method comprises the following steps: acquiring a micro-seismic monitoring signal with a preset duration as a to-be-analyzed signal; in the time domain, extracting a background noise signal as a first signal according to the signal variance ratio in the front and rear time periods of the sliding window; and performing synchronous extrusion wavelet transform on the to-be-analyzed signal to remove noise, and performing inverse transform on the denoised signal wavelet coefficient to obtain a denoised time-domain micro-seismic signal. When the to-be-analyzed signal is subjected to synchronous extrusion wavelet transform, the wavelet coefficient of the to-be-analyzed signal is used as a second wavelet coefficient, the second wavelet coefficient is filtered based on a wavelet coefficient threshold, and the wavelet coefficient threshold is determined based on a probability distribution function of the wavelet coefficient of the first signal. According to the method, the signal-to-noise ratio can be improved, the characteristics of the original vibration signal monitored by microseism are reserved to a great extent, and the influence on subsequent microseism signal processing and analysis is reduced.

The invention provides a synchronous motor rotor winding turn-to-turn short circuit fault diagnosis method based on multi-source information. The synchronous motor rotor winding turn-to-turn short circuit fault diagnosis method comprises the following steps: acquiring exciting current, rotor vibration and stator vibration signals of a synchronous motor working normally and under different rotor winding turn-to-turn short circuit working conditions in real time; converting the one-dimensional time sequence signal into a two-dimensional grayscale image by using a signal-image conversion method; respectively inputting the two-dimensional grayscale images of the excitation current signal, the rotor vibration and the stator vibration into the optimal two-dimensional CNN model for training to obtain a fault diagnosis loss function and diagnosis accuracy of a single signal, and determining basic trust distribution; and performing decision fusion on the fault diagnosis accuracy rates of the three evidence bodies by using a multi-source information fusion algorithm to realize final diagnosis. According to the method, the characteristics of the original signal are reserved to the greatest extent, the influence of manual characteristic extraction is eliminated, and the accuracy of synchronous generator rotor winding turn-to-turn short circuit fault diagnosis is remarkably improved.

The invention discloses a real-time bridge structure damage state recognition method and system. The method comprises the steps of S1, obtaining to-be-tested data; s2, compressing the to-be-tested data at a first compression ratio to obtain first compressed data, classifying the compressed data by using a first classification algorithm to obtain a first classification result, and if the first classification result is damage, entering S3; and S3, compressing the to-be-tested data at a second compression ratio to obtain second compressed data, and classifying the compressed data by using a second classification algorithm to obtain a second classification result. Or the method comprises the steps of S1, S2, S3 and S4; and S4, compressing the to-be-tested data at a third compression ratio, andclassifying the compressed data by using a third classification algorithm to obtain a third classification result, wherein the second compression ratio is greater than the third compression ratio butless than the first compression ratio. Compared with a traditional method, the method has the advantages that massive high-dimensional bridge monitoring time series data can be more efficiently, flexibly and better analyzed, and the bridge damage state is evaluated in real time.

The invention discloses an automatic detection and identification method and detection and identification system grey green fine foreign fibers, and the method comprises the steps: dynamically collecting a real-time cotton flow image of an image collection region, and processing the cotton flow image into an RGB true color image; segmenting the RGB true color image according to a set threshold, and marking pixels of a grey green area in the RGB true color image to obtain a different color point area; calibrating pixels with gradient changes having special-shaped characteristics in the RGB truecolor image through a gradient calculation method, and obtaining a special-shaped area; cleaning and filtering intersection pixels of the heterochromatic area and the special-shaped area, and removing noise points; and determining the position of the foreign fiber in the filtered image by adopting a sliding mask mode, and performing marking to finish foreign fiber identification. According to theautomatic detection and identification method and the detection and identification system, the grey green foreign fiber identification rate can be improved, the false spraying rate of normal cotton is reduced, and the quality of final cotton products of a textile mill is improved.