67 results about "Consecutive frame" patented technology

The invention provides a video compression method, a video decompression method and related devices. The video compression method comprises the following steps: acquiring a space-time block to be processed, wherein the space-time block comprises a plurality of continuous frames; a feature extraction module in an encoder is used to extract spatial-temporal features of the to-be-processed spatial-temporal block, and the encoder is constructed through a 3D convolutional neural network; and performing optimization processing on the spatio-temporal features through a spatio-temporal attention module in the encoder to obtain a coded representation of the to-be-processed spatio-temporal block, the coded representation being capable of being used for decoding and reconstructing the to-be-processed spatio-temporal block. According to the method, the time and space correlation in multiple continuous frames can be utilized at the same time, so that the space-time redundant information of the video is effectively captured and compressed, the capability of capturing global features of the video is improved, and a space-time attention mechanism not only enhances the sensitivity of the network to important features, but also improves the accuracy of the network. And the accuracy and the resolution of feature representation are greatly improved.

In multimedia object tracking and merging of tracked objects, an object is tracked through frames of multimedia content until a frame appears in which the tracked object is not detected. A first track is designated as one or more consecutive frames in which the tracked object is detected, the first track ending at the first frame. Tracking continues to try to detect the tracked object in a second frame subsequent to the first frame. If the tracked object is not again detected, information about the first track is output. If the tracked object is detected subsequently, a second track of consecutive tracked object detection is designated. The tracked objects in the two tracks are then compared with the aid of trained data models, and a matching score is determined to reflect the degree of match. If the matching score meets or exceeds a first threshold, the compared tracks are merged using the same identifier assigned to both tracks. If the matching score does not exceed a second threshold that is less than the first threshold, the tracks may be discarded as showing no match. If the matching score falls between the first and second thresholds, an indication is output for further analysis of the compared tracked objects.

An encoder and method for encoding a media signal is presented. The encoder is configured to perform a first-pass encoding of the media signal with restricting the first-pass encoding onto a set of frames out of a sequence of immediately consecutive frames of the media signal and using a probe quantization parameter, QP, for each of the set of frames of the media signal, so as to obtain a first-pass bitrate for each of the set of frames, determine, based on the first-pass bitrate for each of the set of frames, a start QP for each of the sequence of immediately consecutive frames; and perform a second-pass encoding of the media signal with using the start QP for each of the sequence of immediately consecutive frames, wherein the sequence of immediately consecutive frames is a group of pictures having a hierarchical referencing structure with temporal layers including a temporal base layer 0 up to a highest temporal layer N-1 and the encoder is configured to select the set of frames out of the sequence of immediately consecutive frames so that the set of frames includes all frames of temporal layers 0 to k-1, and, for each temporal layer k to N-1, only a proper subset of the frames of the sequence of immediately consecutive frames which belong to the respective temporal layer, wherein 1 < k < N, and / or the encoder is configured to perform scene detection to detect scene changes, and select the set of frames out of the sequence of immediately consecutive frames depending on whether any of the scene changes falls into the sequence of immediately consecutive frames so that the set of frames represents a sequentially subsampled subset of a sequence of immediately consecutive frames of the media signal in case of none of the scene changes falling into the sequence of immediately consecutive frames, and / or the encoder is configured to perform scene detection to detect scene changes separating scenes, and determine, based on the first-pass bitrate for each of the set of frames, the start QP for each of the sequence of immediately consecutive frames in a manner depending on the scene changes so that, for each of the sequence of immediately consecutive frames, the start QP is exclusively determined based on the first-pass bitrate of one or more frames within the set of frames, which fall into a scene into which the respective frame falls.

Embodiments of the present application disclose a video frame interpolation method. The method comprises: acquiring a plurality of consecutive frames from a video file to be processed, the plurality of consecutive frames comprising a target frame and the frame preceding the target frame and the frame following the target frame; on the basis of an optical flow estimation model, obtaining a reverse optical flow map between the target frame and the preceding frame and a forward optical flow map between the target frame and the following frame; on the basis of the similarity between the forward optical flow map and the reverse optical flow map, determining whether a transition exists among the plurality of consecutive frames; and, when no transition exists among the plurality of consecutive frames, interpolating a frame between the target frame and the following frame by means of an optical flow frame interpolation model. The embodiments of the present application further disclose a video frame interpolation system, a computer device, and a computer storage medium. The technical solution provided by the embodiments of the present application can prevent the appearance of artifacts in a synthesized intermediate frame due to the use of an unsuitable optical flow frame interpolation model for frame interpolation when a transition exists, while also avoiding misjudgment of a transition scene.

Systems and methods for universal phase recognition for intraoperative and postoperative applications are provided. The system receives a video stream that captures a procedure over a time interval with a robotic medical system. The system generates, from the video stream, sets of consecutive frames. A first set of consecutive frames can include a first temporal resolution, and a second set of consecutive frames can include a second temporal resolution. The system determines, via the sets of consecutive frames input into a first model trained with machine learning, phases of the procedure on a moment-to-moment basis over the time interval. The system inputs the phases of the procedure into a second model to generate phase segment over the time interval. The second model can be trained with machine learning based on historical workflows. The system provides an action based on a metric of the phase segment.

This information processing device makes inferences (e.g., object detection) by decimating consecutive frames on a prescribed interval. Upon detecting an event by inference with respect to a frame, the information processing device makes inferences with respect to consecutive frames to the frame in which the event was detected. If the event is no longer detected as a result of the inferences with respect to the frames, the information processing device again makes inferences by decimating frames on the prescribed interval. The information processing device also makes inferences by going back to a frame preceding the frame in which the event was detected.

The invention discloses a multi-frame association long-distance dynamic small target identification method based on frequency domain enhancement, which relates to the field of image processing, and is characterized in that a frequency domain imaging principle and a multi-frame association method are combined, and a time domain image is converted to a frequency domain by using an image sequence acquired by a single photon counting camera Photon Force PF32, so that a target identification result is obtained; the method comprises the following steps: effectively distinguishing the spectrum characteristics of a small target from the spectrum characteristics of noise and background interference, suppressing the noise and background interference through Fourier transform, introducing a multi-frame association method, and analyzing a multi-frame image after frequency domain processing to obtain a motion track and a characteristic change rule of the small target between continuous frames; therefore, a clearer target is obtained through shift accumulation. Through the method, the advantage of high sensitivity of the single-photon counting camera is brought into full play, the problem that the false detection probability of the long-distance dynamic weak and small target is large is effectively solved, and the accuracy and reliability of target detection are improved.

Techniques related to performing object or person association or correspondence in multi-view video are discussed. Such techniques include determining correspondences at a particular time instance based on separately optimizing correspondence sub-matrices for distance sub-matrices based on two-way minimum distance pairs between frame pairs, generating and fusing tracklets across time instances, and adjusting correspondence, after such tracklet processing, via elimination of outlier object positions and rearrangement of object correspondence.

The invention provides a continuous frame parameter adjustment method, electronic equipment and a system in on-board diagnostic communication. The method comprises the following steps: determining the priority of UDS service; determining a reference value of the first parameter according to the priority of the UDS service; determining a predicted value of a network load at a future moment through a network load prediction model based on the correlation characteristics of the vehicle at the current moment; and correcting the reference value of the first parameter according to the predicted value of the network load at the future moment and the true value of the network load at the current moment to obtain a corrected value of the first parameter. Through the technical scheme of the invention, the continuous frame parameter can adapt to the dynamically fluctuating network load, the loss probability of the continuous frame is reduced, and the stable and effective transmission of the continuous frame is ensured.

The invention provides a touch action recognition method and device, a storage medium and electronic equipment, and belongs to the technical field of human-computer interaction. The method comprises the following steps: acquiring multi-frame queue data; calculating an effective contact number of the current frame data in the multi-frame queue data, and determining an action type to which the touch action of the current frame belongs based on the effective contact number; if the action type is a point contact type, calculating a corresponding dispersion degree according to the current frame data and the previous frame data, and determining whether the touch action of the current frame is one of a light touch action or a sliding action based on the dispersion degree; if the action type is a beating type or a palm type, calculating the number of continuous frames which are judged to be the same action type in the multi-frame queue data; and determining whether the touch action of the corresponding frame is one of a slow shooting action, a snapshot action, a long press action, a continuous touch action or a single touch action based on the number of the continuous frames. According to the invention, the recognition accuracy of complex gestures can be improved.

Systems and methods are provided for one or more processors; and memory storing instructions that, when executed by the one or more processors, cause the system to perform: receiving successive frames of sensor data, the successive frames comprising a first frame and a second frame; determining transformations, in sensor coordinates, between coordinates of corresponding elements in the successive frames; determining a mapping between the transformations in sensor coordinates and transformations in geospatial coordinates of the corresponding elements in the successive frames; and determining second geospatial coordinates of the corresponding elements of a third frame based on: a transformation between the second frame and the third frame, and the mapping.

A method for compressing a video stream includes retrieving a plurality of frames corresponding to the video stream. For each of two or more sequential frames of the plurality of frames of the video stream, the method includes extracting Key Point Descriptors (KPDs) for the respective frame and processing the respective frame using Principle Component Analysis (PCA) followed by vector quantization, resulting in a quantized explained variance matrix for the respective frame. The quantized explained variance matrix for the respective frame is stored. The KPDs for the respective frame are stored.

A sequence of frames is encoded in the following manner. An evaluation on a frame-by-frame basis is made for at least a zone in a series of successive frames whether the zone should be encoded in an intra-mode or in an inter-mode. A refresh scheme is applied that frequently causes the zone to be forcibly encoded in the intra-mode for one out of several successive frames, irrespective of the evaluation made. The zone in the series of successive frames is quantized so that if, for a frame, the zone is forcibly encoded in the intra-mode and contrary to the evaluation that was made, the zone is quantized with a higher resolution than if the zone were encoded in accordance with the evaluation that was made in one of the following modes: the intra-mode and the inter-mode.

Approaches are disclosed for enhancing frame rate and visual smoothness in real-time video streams through multi-frame interpolation. A classification neural network analyzes two sequential frames, outputting confidence scores that indicate the reliability of motion data for each pixel. These scores determine whether a pixel's motion is accurately described by motion vectors or should be treated as static. The classification results are reused to generate intermediate frames by warping the original frames based on the motion characteristics. Blending weights are calculated by combining warped motion vector confidence values with static values, and a second neural network refines the alignment and blending of candidate frames. This second network predicts intermediate flows and generates new blending weights, which are used to warp and blend the candidate frames, ultimately producing a final interpolated frame that enhances visual smoothness and consistency in the video stream.

A method of video evaluation for a video comprising successive video frames comprises the steps of generating optical flow vectors for at least a subset of corresponding pixels between frame pairs, calculating a difference between corresponding optical flow vectors of successive frame pairs, and evaluating the resulting differences with respect to a predetermined criterion, to identify respective frame pairs for further processing.

In one embodiment, a computing device receives, from one or more cameras, a video stream comprising multiple frames, where the video stream is received at a first quality. The computing device analyzes, using a machine-learning model, images in the frames, where the machine-learning model has been trained to detect one or more objects-of-interest in the images. The computing device identifies a sequence-of-interest including consecutive frames of the video stream, where at least one object-of-interest was detected in at least one of the consecutive frames. The computing device generates a video package including the sequence-of-interest.

The present invention provides a method and device for removing speed blurred false targets, the method comprising: obtaining temporary tracks of continuous Q frames and target point tracks associated with the temporary tracks in each frame, where Q is a positive integer and Q≥2; calculating the distance R between the target point tracks. (i，j) , i and j represent the frame number, i, j∈{1, 2, 3, ..., Q}, i<j; based on the distance R (i，j) and the distance R (i，j) The interval time T (i，j) Calculate multiple distance change rates V (i，j) For multiple distance change rates V (i，j) Data fusion is performed to obtain the true range change rate corresponding to the temporary track; the true range change rate is compared with the defuzzification speed parameter to obtain a speed difference; the speed difference is compared with a preset difference threshold parameter, and based on the comparison result, it is determined whether to upgrade the temporary track to a mature track. The present invention can identify and eliminate false targets with defuzzification errors, avoiding false target output and erroneous upgrade of temporary tracks.

A method for compressing a video stream includes retrieving a plurality of frames corresponding to the video stream. For each of two or more sequential frames of the plurality of frames of the video stream, the method includes extracting Key Point Descriptors (KPDs) for the respective frame and processing the respective frame using Principle Component Analysis (PCA) followed by vector quantization, resulting in a quantized explained variance matrix for the respective frame. The quantized explained variance matrix for the respective frame is stored. The KPDs for the respective frame are stored.

This invention relates to a method for tracking moving ship targets in remote sensing images, applied to remote sensing scenarios with a large field of view and low frame rate. The method compensates for the limitations of single-frame detection features by combining spatial information of the marine ship target with local details of its motion and multi-frame correlation. The process includes continuously acquiring multiple frames of remote sensing images and selecting ship targets to be tracked; determining the identity of the selected ship targets in two consecutive frames; and generating the trajectory of the moving ship target based on its position in the multi-frame images. This achieves high detection performance with relatively low complexity.

Methods and systems of frame based image segmentation are provided. For example, a method for feature object tracking between frames of video data is provided. The method comprises receiving a first frame of video data, extracting a mask feature for each of one or more objects of the first frame, adjusting the first frame by applying each initial mask and corresponding identification to a respective object of the first frame, and outputting the adjusted first frame. The method further comprises tracking the one or more objects in one or more consecutive frames. The tracking comprises extracting a masked feature for each of one or more objects in the consecutive frame, adjusting the consecutive frame by applying each initial mask and corresponding identification for the consecutive frame to the respective object of the one or more objects of the consecutive frame, and outputting the adjusted consecutive frame.

The invention discloses a threshold-adaptive three-frame difference detection algorithm. The algorithm comprises the following specific steps: S1, carrying out alignment operation on three frames of images to be processed; s2, carrying out difference operation on the aligned two continuous frames of images; s3, selecting a maximum value of the difference image, and preliminarily determining a target position; s4, finally determining the number N of the targets; s5, performing target detection operation by using the threshold value to obtain an optimal detection threshold value; and S6, obtaining a final detection target position by using the updated optimal detection threshold value Tmin. And S7, repeating the operations from S1 to S6 on all the to-be-processed images according to the time sequence until the task is finished. The method solves the problem that a double-frame difference detection fixed threshold algorithm does not have real-time performance, robustness and accuracy, is not only suitable for the target detection field such as infrared detection, but also can be migrated to other complex target detection fields needing high precision, high timeliness and the like, and has wide application prospects.

This invention provides a system and method for providing video semantic segmentation. Semantic segmentation is performed on a first frame in a sequence of video frames to obtain at least one first semantic feature of the first frame. Semantic segmentation is performed on a second frame in the sequence to obtain at least one second semantic feature of the second frame, wherein the second frame follows the first frame. Semantic segmentation is performed on a third frame in the sequence to obtain at least one third semantic feature, wherein the third frame follows and is also following the first frame by a first predetermined number of consecutive frames. At least one first semantic feature, at least one second semantic feature, and at least one third semantic feature are combined to form at least one fourth semantic feature of the second frame.

The application relates to a video translation method and device, which comprises the following steps: inputting image features and coordinate features of original data into a fusion input generator; calculating a three-frame loss according to continuous frames of a video of the input generator, and optimizing the generator according to the three-frame loss; and performing video translation based on the optimized generator, so as to improve the visual effect of the converted video.

Disclosed are an encoder for encoding a video sequence having successive frames into a bitstream, a method for encoding a video sequence having successive frames into a bitstream, a decoder for decoding the bitstream to reconstruct the video sequence having successive frames, and a method for decoding the bitstream to reconstruct the video sequence having successive frames.

The present invention provides an ultrasonic object volume calculation method and an ultrasonic system, which include: collecting a plurality of continuous frames of an object at a plurality of time points by an ultrasonic reflection method; detecting the object in each of the plurality of frames; when the first frame of the plurality of frames is detected to have the object, determining the predicted contour, the minimum circumscribed quadrilateral and the maximum inscribed quadrilateral of the object in the first frame; determining the contour of the object according to the predicted contour, the minimum circumscribed quadrilateral and the maximum inscribed quadrilateral of the object in the first frame; and converting the object into a three-dimensional space according to the contour of the object to calculate the volume of the object; wherein the plurality of frames are two-dimensional images. The ultrasonic system and the ultrasonic object volume calculation method of the present invention can segment the object in real time to display it on a display, and quickly calculate the volume of the object.

To generate communication specifications of a device whose communication specifications are unknown.SOLUTION: A communication specification generation device (101) includes a frame extraction unit (11) that extracts a frame having a specific frame length from communication log data generated by a target device (103), a header identification unit (12) that sequentially compares values in a range determined by a standard header length from a head of the frame between two consecutive frames to identify a header of the frame, and a combination identification unit (13) that extracts codes of a command and a response from the header on the basis of position information and size information about the respective codes of the command and response frames and identifies a combination of the codes of the command and the response according to a communication state of the command and response frames in the communication log data.SELECTED DRAWING: Figure 1

The invention discloses a detection method and device, electronic equipment, a storage medium and a computer program product. The method comprises: determining M segments of frame sequences from a video to be detected, M being an integer greater than or equal to 1, each segment of frame sequence in the M segments of frame sequences including N continuous frames, the N continuous frames included in each segment of frame sequence being completely different, and N being an integer greater than 1; for each section of frame sequence in the M sections of frame sequences, performing mask processing on N continuous frames contained in the frame sequence in the to-be-detected video to obtain a mask video; by using data contained in the mask video, sequentially regenerating N continuous frames to be masked according to a sequence from two ends to the middle; detecting the frame sequence by using N continuous frames contained in the frame sequence and the regenerated N continuous frames to obtain a detection result corresponding to the frame sequence; and determining whether the to-be-detected video belongs to a forged video or not by using all the detection results corresponding to the M segments of frame sequences to obtain a first detection result.

The invention relates to a flying dust detection method and device, electronic equipment and a readable storage medium. The method comprises the following steps: respectively acquiring a radar point cloud and a depth map of a current frame through a radar and a depth camera; wherein the radar point cloud and the depth map of the same frame correspond to each other and form a parameter group; updating a frame queue of the current frame based on the parameter group, the frame queue being formed by a plurality of parameter groups corresponding to continuous frames, and the frame queue comprising the parameter group of the current frame; and determining a fusion score of the parameter group of the current frame, and determining a target detection result of the current frame based on the fusion scores of all the parameter groups in the frame queue. By adopting the method, the dust raising environment can be accurately detected.

This disclosure relates to multi-frame interpolation for real-time video processing using deep neural networks. A method is disclosed to improve frame rate and visual smoothness in real-time video streams through multi-frame interpolation. A classification neural network analyzes two consecutive frames and outputs confidence scores indicating the reliability of motion data for each pixel. These scores determine whether the motion of a pixel is accurately described by a motion vector or should be considered static. The classification results are reused to generate intermediate frames by warping the original frames based on motion features. Blending weights are calculated by combining the warped motion vector confidence values ​​with static values, and a second neural network refines the alignment and blending of candidate frames. This second network predicts the intermediate stream and generates new blending weights, which are used to warp and blend candidate frames, ultimately producing final interpolated frames that enhance the visual smoothness and consistency of the video stream.

The invention provides a point cloud labeling method and device, equipment and a medium, and relates to the field of image processing, in particular to the technical field of computer vision, artificial intelligence, automatic driving and intelligent traffic. The specific implementation scheme is as follows: obtaining point cloud data of a first frame in continuous multi-frame point cloud data; determining a labeling type according to the obtained target feature of the first frame; the annotation type comprises a superimposed frame annotation, a predicted frame annotation or a single frame annotation; according to the labeling type, determining to-be-labeled point cloud data; determining a labeling operation according to the labeling type; and in the to-be-labeled point cloud data, based on the labeling operation, determining a labeling result of the selected point. According to the embodiment of the invention, the point cloud labeling efficiency and accuracy can be improved.