29 results about "Interframe correlation" patented technology

The invention relates to the technical field of streaming media decoding, and discloses a streaming media decoding method and device for a network high-definition player and a storage medium, and the method comprises the steps: carrying out the multi-thread parallel collection and storage of real-time temperature data, network bandwidth jitter data and video code rate fluctuation data of a decoding chip in the network high-definition player; obtaining a streaming media dynamic feature data set; performing classification training to obtain an inter-frame correlation prediction model; constructing a multi-stage streaming media caching system based on the inter-frame correlation prediction model; monitoring a cache state, a network state and a chip state in the multi-stage streaming media cache system in real time to obtain a decoding event sequence; according to the decoding event sequence, the decoding parameter is adjusted in real time, and the streaming media decoding control instruction is obtained.The collaborative optimization of a network layer, a decoding layer and a display layer is achieved, the playing starting delay is reduced, and the overall stability of a network high-definition player is improved.

According to the method, the system, the medium and the equipment for monitoring the abnormal state of the wind generating set, a self-adaptive noise reduction mechanism, feature extraction and model classification are combined, so that physical information features are effectively extracted, and accurate state detection is achieved. The method comprises the following steps: firstly, building a signal acquisition system, acquiring an acoustic signal in real time, and executing direct current offset removal preprocessing to eliminate low-frequency noise; secondly, noise reduction of the original sound signal is completed by using a self-adaptive noise reduction method; then, extracting the MFCC characteristics of the noise reduction signal; then, calculating an MFCC inter-frame correlation feature and an aerodynamic-harmonic imbalance factor, and forming a fusion feature vector; and finally, constructing a physical neural network model based on physical information to carry out abnormal state detection.

One or more methods, apparatuses, computer-readable storage mediums, and systems for entropy coding vertex information of an edge in a voxelized space of a point cloud are disclosed. Symbols of a neighborhood configuration of a current edge may be determined based on one or more already coded edges. The already coded edges may be selected from a spatial topology of edges. The use of a motion-compensated point cloud for coding centroid residual values may enhance interframe correlation used for determining a context or probability model. This increased correlation may improve the selection of coders, leading to enhanced compression of centroid residual values.

The application provides a compression artifact suppression method combined with multi-level interframe correlation. The method mainly relates to a means combining multi-level motion compensation and deep convolution mapping to perform quality enhancement on HEVC decoded video. The application mainly utilizes the characteristics of interframe correlation of the video, and then designs a multi-level motion compensation network to perform motion compensation on two adjacent frames of a current frame, and then fuses features of the current frame with the two adjacent frames. Then, attention feature extraction is performed on the current frame and the fused multi-frame feature map respectively. Finally, deep residual learning and cross-layer cascade are performed to realize deep fusion, mapping and reconstruction of multi-level features, so that quality enhancement of the current frame is realized.

The present invention relates to the technical field of streaming media decoding, and discloses a streaming media decoding method, device and storage medium for a network high-definition player. The method includes: performing multi-threaded parallel acquisition and storage on the real-time temperature data, network bandwidth jitter data, and video bitrate fluctuation data of the decoding chip in the network high-definition player to obtain a streaming media dynamic feature data set; performing classification training to obtain an inter-frame correlation prediction model; constructing a multi-level streaming media cache system based on the inter-frame correlation prediction model; performing real-time monitoring on the cache state, network state, and chip state in the multi-level streaming media cache system to obtain a decoding event sequence; and performing real-time adjustment on the decoding parameters according to the decoding event sequence to obtain a streaming media decoding control instruction. The present invention realizes the collaborative optimization of the network layer, decoding layer, and display layer, reduces the playback startup delay, and improves the overall stability of the network high-definition player.

The invention discloses a robust sound source localization method fusing sound intensity amplitude inter-frame correlation, and the method comprises the steps: receiving a time domain voice signal through a regular triangle array, and carrying out the short-time Fourier transform of the time domain signal, and obtaining a time-frequency domain signal; calculating a sound intensity component by using a sound intensity estimation method of sound pressure Taylor series expansion; calculating a comprehensive weight of each time frequency point by using a sound intensity estimation consistency algorithm, and calculating a sound intensity amplitude at the same time; weighting the sound intensity amplitude by using the comprehensive weight; constructing a reliable time-frequency point screening factor fusing the inter-frame correlation of the sound intensity amplitudes by utilizing the strong correlation between the weighted sound intensity amplitudes and the unweighted sound intensity amplitudes; screening out time-frequency points which are less influenced by reverberation and noise by using the screening factor, and superposing the sound intensity corresponding to the screened time-frequency points to obtain an azimuth angle estimated value; the method provided by the invention has better anti-reverberation and anti-noise capabilities in indoor sound source localization.

The application provides a coding method and device in AIGC generation. The decoding method in AIGC generation of the application comprises the following steps: obtaining first feature information and motion information of a current frame based on generated feature information output by an AIGC generation model, wherein the motion information is used to represent inter-frame correlation; performing motion modeling according to the first feature information and the motion information of the current frame to obtain second feature information of the current frame, wherein the motion modeling is used to compensate the first feature information of the current frame based on the motion information; and obtaining reconstructed content of the current frame according to the second feature information of the current frame. The application can improve the time domain compression efficiency and the quality of the reconstructed content in AIGC.

The application discloses a video abstracting method based on a transformer and a TCN, and comprises the following steps: performing feature extraction by using a pre-trained convolutional network; extracting time sequence information by using a time sequence convolutional network; inputting spatial feature sequences and time sequence feature sequences of a video into a local transformer module and a global transformer module respectively, calculating the interframe correlation and the long-term dependence between video sequence nodes, and fusing the features; predicting the importance of each frame; dividing the video into segments, calculating the importance of each segment, selecting segments with high importance, and generating a dynamic abstracting. The above process trains a video abstracting model in a supervised manner by using a labeled daily video abstracting data set. The application emphasizes the extraction and separation of interframe correlation and long-term dependence in the video abstracting task on the basis of time sequence modeling, and improves the understanding ability of the model for the video.

Various embodiments include techniques for performing hardware-assisted weighted prediction encoding with software-generated weight prediction parameters for a media frame. The disclosed weighted prediction encoding techniques take advantage of temporal interframe correlation between adjacent media frames in a video stream. In certain examples, a media frame, or portions thereof, can exhibit a rapid luminance change, such as during a fade in from black effect, a fade out to black effect, a video depicting an explosion or blast, and / or the like. This rapid luminance change reduces interframe correlation, even in cases where the interframe correlation would be higher if not for the luminance change. The disclosed techniques remove this luminance change before encoding, resulting in higher temporal correlation between adjacent frames. The higher temporal correlation can result in reduced bit rate and improved visual quality of the resulting output video stream.

The present invention provides a method and device for optimizing time domain context, which obtains reference frames and reconstructs motion vectors and generates predicted frames; extracts the inter-frame correlation between the predicted frame and the reference frame through a motion estimation network, and determines the directional motion vector based on the inter-frame correlation; uses the directional motion vector and the reference frame to perform time domain alignment and feature extraction to generate a directional time domain context; performs feature fusion calculation on the directional time domain context and the propagation time domain context to obtain a compensated time domain context. The present invention is based on an end-to-end video coding framework of conditional coding, and uses reference frames, reference features, and reconstructed motion vectors to perform iterative calculations at the decoding end to fine-tune the time domain context, thereby significantly improving the quality and accuracy of the time domain context. This optimization effectively improves the accuracy of motion compensation and optimizes the entire video coding process, reducing unnecessary data transmission and storage requirements, and ultimately achieving more efficient video coding.

The application discloses a space target trajectory correlation method and system, wherein the system comprises the following steps: obtaining target information detected by a detector in each calculation period; performing interframe trajectory correlation on the target according to the target information in each calculation period to obtain a trajectory correlation result, and storing the trajectory correlation in an interframe correlation target library; and obtaining target information of each frame from the interframe correlation target library, performing trajectory correlation on the target between frames, and obtaining target trajectory information of the current frame. The application effectively solves the problems of insufficient real-time performance, error accumulation and dimension explosion of traditional methods.

The invention provides a robust super-directional beam forming method and system based on a Kronecker product. Inherent inter-frame correlation between adjacent frames of a voice signal in a short-time Fourier transform domain is introduced into robust super-directional beam forming: firstly, a signal model observed by a microphone array is expressed as a Kronecker product form containing a steering vector and an inter-frame correlation vector, and the steering vector is determined by a geometric structure of the array; the inter-frame correlation vector can be obtained by adopting three methods of time-invariant long-time estimation, time-varying recursive adaptive estimation or data-driven deep learning estimation. And then designing a super-directional beam former into a Kronecker product of a time domain filter and a spatial domain filter, optimizing the time domain filter to realize maximization of white noise gain, and optimizing the spatial domain filter to realize maximization of directional gain. Compared with a traditional scheme, the noise suppression robustness is remarkably improved through combination of time-space domain optimization, and better noise reduction performance is shown in a complex acoustic environment.

The application discloses a real-time multi-target detection system and method based on deep learning, and relates to the technical field of deep learning.The application utilizes interframe correlation multiplexing of overlapping area features through a space-time multiplexing module, thereby reducing redundant calculation; a feature enhancement module differentiates processing of dynamic and overlapping stable area features, strengthens basic dynamic features of a motion area, simplifies basic static features of an overlapping stable area, and performs weighted fusion on transition area features; and a window generation module dynamically adjusts detection window parameters based on target density.The application solves the problem that real-time performance and detection accuracy are difficult to be considered in the prior art, and improves processing efficiency and detection accuracy.

The application discloses a three-dimensional geometric model sequence compression method and device based on binary arithmetic coding, equipment and a storage medium, and the method comprises the following steps: segmenting a three-dimensional geometric model sequence based on the three-dimensional model data type of the three-dimensional geometric model sequence, obtaining a plurality of substructures, symbolizing the mesh patches of the three-dimensional geometric model with patch connection attributes, classifying the three-dimensional geometric model with vertex attributes based on interframe correlation, predicting the residual of the three-dimensional geometric model with vertex attributes based on the interframe correlation attributes, binarizing the quantized residual and / or the symbolized mesh patches, obtaining a binary bit string, and encoding the binary bit string, so that the three-dimensional geometric model sequence compression is realized, the multi-dimensional attributes of the dynamic three-dimensional geometric model sequence are effectively processed, the compression ratio and the compression efficiency are greatly improved, and the timeliness requirement of the three-dimensional geometric model sequence is met.

The application discloses a robust sound source positioning method fusing sound intensity amplitude interframe correlation, and comprises the following steps: receiving time domain voice signals by using a regular triangle array, and performing short-time Fourier transform on the time domain signals to obtain time-frequency domain signals; sound intensity components are calculated by using a sound pressure Taylor series expansion sound intensity estimation method; each time-frequency point comprehensive weight is calculated by using a sound intensity estimation consistency algorithm, and sound intensity amplitudes are calculated simultaneously; the sound intensity amplitudes are weighted by using the comprehensive weights; a reliable time-frequency point screening factor fusing sound intensity amplitude interframe correlation is constructed by using the strong correlation between the weighted sound intensity amplitudes and the unweighted sound intensity amplitudes; time-frequency points less affected by reverberation and noise are screened by using the screening factor; the corresponding sound intensities of the screened time-frequency points are superposed to obtain an azimuth angle estimation value; and the method has better anti-reverberation and anti-noise capabilities in indoor sound source positioning.

The present invention discloses a correlation demodulation device applicable to a single-photon detector, comprising: a single-photon detector; a single-photon detector readout circuit, including a first encoding circuit and a first accumulator, wherein the input end of the first encoding circuit is electrically connected to the output end of the single-photon detector, and the output end of the first encoding circuit is electrically connected to the input end of the first accumulator; a non-linear accumulator based on inter-frame correlation, whose input end is electrically connected to the output end of the first accumulator; a histogram statistical algorithm unit based on synchronous sampling, whose input end is electrically connected to the output end of the non-linear accumulator based on inter-frame correlation; a common-mode rejection unit, whose input end is electrically connected to the output end of the histogram statistical algorithm unit based on synchronous sampling, and the output end of the common-mode rejection unit is electrically connected to the input end of the histogram statistical algorithm unit based on synchronous sampling. The present invention can improve the signal-to-noise ratio of the measurement result of the single-photon detector.

The invention relates to the technical field of data compression and storage, and discloses a mass video monitoring high-compression storage system and device based on artificial intelligence, and the system comprises a video data acquisition module, a video data processing module, a video data processing module and a video data storage module, obtaining video frame data from the original video stream by using a video frame collector; the data redundancy analysis module is used for analyzing the correlation coefficient between the video frames and the data redundancy of the video frames; the self-adaptive compression coding module is used for self-adaptively adjusting compression coding parameters according to the redundancy; the quality evaluation and adjustment module is used for carrying out quantitative evaluation on the quality of the video subjected to self-adaptive compression coding, and optimizing the video quality by adjusting quantization parameters when the quality does not reach the standard; and the video output module outputs the compressed video file, the compression quality report, the compression efficiency statistics and the storage space saving report. According to the invention, the video quality is ensured, the high compression rate of the video is realized, and the storage space is saved.

The present invention relates to a deepfake video forensics method based on semi-supervised learning. Based on the semi-supervised learning method, only relying on real videos as input, the detection of deepfake videos is realized. The negative sample generator proposed in the invention effectively simulates the tampering traces left in the last step of deepfake video synthesis. The data augmentation module emphasizes the differential information between genuine and fake images, accelerating the learning process of the detection model at the data level. The lightweight semi-supervised classifier combines RGB images, low-level spatial noise maps, and inter-frame correlated temporal noise maps to find forensics clues at multiple scales. The three complement each other and jointly form an efficient and robust semi-supervised Deepfake video detection framework.

Provided in the present application are a coding method and apparatus during AIGC generation, and a decoding method and apparatus during AIGC generation. The decoding method during AIGC generation in the present application comprises: on the basis of generated feature information outputted by an AIGC generative model, obtaining first feature information of the current frame and motion information, wherein the motion information is used for representing inter-frame correlation; on the basis of the first feature information of the current frame and the motion information, performing motion modeling, so as to obtain second feature information of the current frame, wherein the motion modeling is used for performing compensation on the first feature information of the current frame on the basis of the motion information; and on the basis of the second feature information of the current frame, acquiring reconstructed content of the current frame. The present application can improve the time-domain compression efficiency and the quality of reconstructed content in AIGC.

The invention discloses a multi-channel filtering method for optimizing a full-band signal-to-noise ratio based on inter-frame correlation, and the method specifically comprises the steps: collecting a band correlation signal in a multi-channel voice system, and converting the band correlation signal to a time-frequency domain; generating a multi-channel vector signal containing inter-frame information; carrying out noise reduction processing on the multi-channel noisy voice signal containing the inter-frame information; calculating a correlation matrix of the multi-channel clean voice signal containing the inter-frame information and the multi-channel noise signal containing the inter-frame information; voice dominant and noise dominant subspaces are distinguished according to the sizes of the characteristic values, the voice dominant subspace with the maximum characteristic value is subjected to noise reduction and enhancement, and other noise dominant subspaces are directly set to zero; and forming a full-band filter with the length of 1 * 10. According to the method, the noise reduction stability, robustness and speech enhancement effect of the filter in a complex noise environment are remarkably improved. The invention further discloses a multi-channel filtering system for optimizing the full-band signal-to-noise ratio based on the inter-frame correlation.

The invention provides a time domain context optimization method and device. The method comprises the following steps: acquiring a reference frame and a reconstructed motion vector and generating a prediction frame; extracting inter-frame correlation between the prediction frame and the reference frame through a motion estimation network, and determining a directional motion vector according to the inter-frame correlation; performing time domain alignment and feature extraction by using the directional motion vector and the reference frame to generate a directional time domain context; and performing feature fusion calculation on the directional time domain context and the propagation time domain context to obtain a compensation time domain context. According to the end-to-end video coding framework based on conditional coding, iterative computation is carried out at a decoding end by using the reference frame, the reference feature and the reconstructed motion vector, and the time domain context is finely adjusted, so that the quality and the accuracy of the time domain context are remarkably improved. The optimization effectively improves the accuracy of motion compensation, optimizes the whole video coding process, reduces unnecessary data transmission and storage requirements, and finally realizes more efficient video coding.

One or more methods, devices, computer-readable storage media, and systems are disclosed for entropy encoding edge vertex information in a voxelized space of a point cloud. The symbols of the current edge's neighborhood may be determined based on one or more previously encoded edges. The previously encoded edges may be selected from the edge's spatial topology. The use of motion-compensated point clouds to encode centroid residual values ​​may enhance inter-frame correlation used to determine context or probability models. This increased correlation may improve coder selection and result in enhanced compression of the centroid residual values.

The present application relates to a target tracking method, device, electronic device and storage medium. When performing target tracking on the current frame image in an image sequence, the present application performs inter-frame sampling on the image sequence with an inverse correlation between the sampling density at different positions and the time interval from different positions to the current frame image, and extracts inter-frame correlation information with a high degree of correlation with the current frame image. The inter-frame correlation information with a high degree of correlation can guide or assist the current frame image to perform more accurate / precise positioning of the target; and further perform intra-frame multi-scale sampling on each of the inter-frame images sampled in the above manner, which can obtain the multi-scale distribution information of the target in the intra-frame image, improve the understanding of a single image frame, and overcome the situation such as the loss of the tracked target easily caused by the fixed receptive field of the convolution kernel in the convolutional neural network.

The present invention provides a video monitoring system and method for a substation. The system includes an image acquisition device, a monitoring main end and a monitoring mobile end, and the image acquisition device and the monitoring mobile end are both communicatively connected to the monitoring main end; the method includes: obtaining video image data of the monitored area in real time, selecting a preset number of frames in the preprocessed image as selected frames, and generating a video sequence with the selected frames in units of frames; performing de-jitter processing on the video sequence through the inter-frame correlation coefficient; performing moving object detection on the video sequence after de-jitter processing, and when a moving object is detected in the video, prompting through an information display module that there is a moving object in the corresponding video, and at the same time sending the information of the appearance of the moving object to the monitoring mobile end; through the system and method, it is possible to maintain a good recognition rate of moving objects even when the background environment of the video image changes.

The present invention belongs to the fields of infrared remote sensing and infrared space technology, and discloses a space-based infrared air target detection method, a storage medium, and a computer device, and determines a reference frame I b and a reference frame I b+l , performs local normalization processing on the b frame, performs local normalization processing on the (b + j)th frame, and performs inter-frame correlation difference; after the inter-frame correlation difference is completed, performs dipole enhancement and spatial difference; multiplies the data after spatial background suppression, dipole enhancement, and spatial difference to obtain an intermediate result; performs global normalization processing, local feature contrast, and adaptive threshold segmentation on the intermediate result data, and performs detection results. The present invention realizes space-based infrared air target detection by constructing a local inter-frame matching difference model and a spatial local contrast model, reduces the processing complexity of infrared air target detection and the resource requirements for hardware implementation, and effectively improves the efficiency of target detection.

The invention provides a video frame interpolation positioning method and system based on graph convolution multi-level inter-frame correlation, relates to the technical field of computer vision, and aims to solve the problems that an existing video frame interpolation detection technology does not consider unnatural traces of forged video frames, processing of time continuity features is insufficient, and video tampering detection precision is low. Comprising the following steps: acquiring three continuous video frames, respectively inputting the three continuous video frames into three node embedding modules, extracting intra-frame features, and acquiring inter-frame relationship representation through an edge embedding module; the method comprises the following steps: constructing a graph structure by taking intra-frame features as nodes and taking relation expressions as edges, updating node states through a graph neural network feature transmission module, obtaining global feature graphs of each frame, splicing the global feature graphs of each frame, inputting the spliced global feature graphs into a depth feature extraction and classification module, and performing feature extraction and classification to obtain a video frame interpolation position. According to the method, the tampering features can be better extracted, and the tampered video frame position can be more accurately positioned.

A video decoding method includes decoding an image bitstream associated with a video sequence, wherein the decoding of the image bitstream reconstructs a key reference frame; factorizing the reconstructed key reference frame into a key frame latent feature and a first group of compact motion vectors associated with the reconstructed key reference frame; decoding a feature bitstream associated with the video sequence to obtain a second group of compact motion vectors associated with an inter frame; transforming, based on the first group and second group of compact motion vectors, the key frame latent feature into a first fine-grained motion field for the reconstructed key reference frame and a second fine-grained motion field for the inter frame; predicting a dense motion based on the first and second fine-grained motion fields; and generating the inter frame based on the dense motion and the reconstructed key reference frame.

The invention discloses a real-time multi-target detection system and a real-time multi-target detection method based on deep learning, and relates to the technical field of deep learning. According to the real-time multi-target detection system and the real-time multi-target detection method based on deep learning, overlapping region features are multiplexed by utilizing inter-frame correlation through a space-time multiplexing module, and redundant calculation is reduced; the feature enhancement module carries out differentiation processing on dynamic and overlapping stable region features, reinforces basic dynamic features of a motion region, simplifies basic static features of an overlapping stable region, and carries out weighted fusion on transition region features; the window generation module dynamically adjusts detection window parameters based on the target density. According to the invention, the problem that real-time performance and detection precision are difficult to consider in the prior art is solved, and the processing efficiency and the detection accuracy are improved.

The invention discloses a nondestructive storage method of a fluorescence spectrum binary concentration image sequence based on multichannel bit mapping and entropy coding. Firstly, N frames of binary edge images with the same resolution are fused bit by bit according to pixels to form a fusion matrix V; and determining the storage bit width of the V, carrying out run length coding on the V, constructing a Huffman code table based on the statistical probability, carrying out entropy compression, and finally generating a compressed file. The header of the file records information such as the image size, the frame number, the data type and the CRC check code, and the integrity and the decompressibility of compressed data are ensured. And during decompression, performing Huffman decoding, inverse RLE and bitwise disassembly according to an inverted sequence to realize complete lossless recovery of an original sequence. By using the method, redundancy among multiple frames of images can be fully utilized, the compression ratio is improved, lossless reconstruction of original data is realized, and the method is particularly suitable for a binary image sequence scene with strong inter-frame correlation. In a multi-frame related scene, the compression ratio of the method is obviously superior to that of a traditional frame-by-frame PNG method, and the reproduction degree reaches 100%.