38 results about "Motion boundary" patented technology

The methods and systems of the present invention enable the estimation of optical flow by performing color segmentation and adaptive bilateral filtering to regularize the flow field to achieve a more accurate flow field estimation. After creating pyramid models for two input image frames, color segmentation is performed. Next, starting from a top level of the pyramid, additive flow vectors are iteratively estimated between the reference frames by a process including occlusion detection, wherein the symmetric property of backward and forward flow is enforced for the non-occluded regions. Next, a final estimated optical flow field is generated by expanding the current pyramid level to the next lower level and the repeating the process until the lowest level is reached. This approach not only generates efficient spatial-coherent flow fields, but also accurately locates flow discontinuities along the motion boundaries.

A method and apparatus for deriving fine-grained motion compensated prediction of boundary pixels in a video coding system are disclosed. Embodiments of the present invention determine one or more neighboring coding units (CUs) adjacent to a current coding unit (CU). For each neighboring CU, motion-compensated prediction is derived for each neighboring CU using the MV of the neighboring CU. The pre-generated predictors at a bottom side or a right side of each neighboring CUs are derived and stored on a smallest CU (SCU) basis. The pre-generated predictors and the motion compensated predictor for a current boundary pixel are combined using weighting factors to form a final predictor for the current pixel.

According to one embodiment, a video encoding apparatus includes a motion vector calculation module, a motion boundary detection module, and a quantization parameter setting module. The motion vector calculation module calculates motion vectors for respective blocks in an input image. The motion boundary detection module detects a motion boundary between a motion region and a still region in the input image based on the motion vectors of the respective blocks. The quantization parameter setting module sets a value of a quantization parameter used to quantize a block that borders the detected motion boundary to a value smaller than a value which is set when the block does not border the motion boundary in accordance with a detection result of the motion boundary.

The invention discloses a human motion identification method based on dense sampling of the motion boundary and a motion gradient histogram. The method mainly comprises steps that 1), a video stream is inputted; 2), the optical flow field of the inputted video is calculated, feature point sampling is carried out, and dense feature points are extracted; 3), the trajectory of the feature points is calculated; 4), dense descriptors along the feature point trajectory are calculated; 5), two adjacent frames of video images are derived in time to obtain time series motion images, and the spatial gradient of the motion images is calculated to obtain a motion gradient descriptor HMG; 6), feature encoding is performed separately for each descriptor; 7), after regularization of each descriptor, thedense descriptors and the motion gradient descriptor are connected in series to form a feature vector; 8), the feature vector is trained and learned to obtain a human motion identification model; and9), the human body motion is identified through utilizing the human motion identification model. The method is advantaged in that motion identification precision is improved, and calculation cost is further reduced.

The invention discloses an adaptively-adjusted action boundary based single phase grounding fault relay protection method for power transmission line. According to the method, a fault phase voltage, a fault phase current, a fault phase negative sequence current and a zero sequence current in the protection installation position of the power transmission line are measured firstly; action boundary adjustment impedance is calculated; the summation of the action boundary adjustment impedance and protection set impedance is calculated to obtain adjusted action boundary impedance; measurement impedance is obtained by dividing the zero sequence compensation current by the fault phase voltage; then whether the measurement impedance amplitude is lower than the adjusted motion boundary impendence amplitude or not is determined; and if the measurement impedance amplitude is lower than the adjusted motion boundary impendence amplitude, a single phase grounding fault occurring in the protection set range of the power transmission line is determined, and a protection apparatus sends out an action tripping signal. The method adopts lumped parameters for modeling, so that simple algorithm principle, easy implementation of the programs, low calculated amount and high protection action speed are realized; the protection set impedance is adjusted in real time by the action boundary impedance, so that the protection action performance is not affected by the transition resistance and the load current; and therefore, the method can carry out correct and reliable action when the power transmission line is in the singe phase high resistance grounding fault.

A method and apparatus for occlusion area detection based on block difference associated with a motion vector and a predicted block difference are disclosed. For each current block of a frame, motion estimation is performed based on a temporally previous frame and a temporally subsequent frame. Based on the motion vector derived, two reference blocks of the current block are located in the temporally neighboring frames. The block difference between these two reference blocks is calculated for the current block. By comparing the block difference with a predicted block difference of the current block, the current block is determined to be an occlusion block or not accordingly. The predicted block difference is updated by averaging the block difference of neighboring blocks in a non-motion boundary area.

The invention discloses a human body behavior identification method and system based on multi-feature linear time sequence coding, and the method comprises the steps: firstly, uniformly dividing an original video into a plurality of sub-segments, constructing a continuous frame set in each sub-segment, and respectively calculating the optical flow between two adjacent image frames in each sub-segment continuous frame set, and obtaining an optical flow image set; Calculating a motion boundary value to obtain a motion boundary image set; Taking a representative frame, an optical flow image set and a motion boundary image set of the original video as input, and obtaining CNN characteristics of each sub-segment of the original video by adopting a CNN network; Performing linear time sequence coding according to the CNN features of different modalities; Respectively carrying out action recognition by adopting an action classification algorithm according to the linear time sequence coding characteristics; And carrying out weighted fusion on the plurality of behavior recognition results to obtain a final behavior recognition result. According to the method, the important behavior representation information of the motion boundary graph is fused, and the integrated learning process of the behavior convolution features and the time sequence features is fused through the linear time sequence coding method, so that the human body behavior recognition accuracy is improved.

A method and apparatus for deriving fine-grained motion compensated prediction of boundary pixels in a video coding system are disclosed. Embodiments of the present invention determine one or more neighboring coding units (CUs) adjacent to a current coding unit (CU). For each neighboring CU, motion-compensated prediction is derived for each neighboring CU using the MV of the neighboring CU. The pre-generated predictors at a bottom side or a right side of each neighboring CUs are derived and stored on a smallest CU (SCU) basis. The pre-generated predictors and the motion compensated predictor for a current boundary pixel are combined using weighting factors to form a final predictor for the current pixel.

The invention relates to an image optical flow optimization method and system based on a residual field and a displacement field. The method comprises the following steps: acquiring a reference image; estimating the reference image by adopting an optical flow estimation method to obtain an initial optical flow field; obtaining an optical flow optimization model, wherein the optical flow optimization model comprises an image encoder, an optical flow encoder, a first decoder and a sub-decoder; inputting the reference image and the initial optical flow field into the optical flow optimization model, and outputting a residual field and a displacement field; superposing the initial optical flow field and the residual field to obtain a preliminarily optimized optical flow field; and resampling the preliminarily optimized optical flow field by using the displacement field to obtain an optimized optical flow field. According to the invention, the precision and robustness of the image sequence optical flow estimation in the motion boundary region can be improved.

A method and apparatus for deriving motion compensated prediction for boundary pixels in a video coding system are disclosed. Embodiments of the present invention determine a current motion vector (MV) and one or more neighboring MVs corresponding to an above MV, a left MV, or both the above MV and the left MV. A first predictor for a current boundary pixel in a boundary region of the current CU is generated by applying motion compensation based on the current MV and a first reference picture pointed by the current MV. One or more second predictors for the current boundary pixel are generated by applying the motion compensation based on the neighboring MVs and reference pictures pointed by the neighboring MVs. A current boundary pixel predictor for the current boundary pixel is then generated using a weighted sum of the first predictor and the second predictors according to weighting factors.

The application provides a vehicle control method, a chassis domain controller, and an intelligent driving vehicle. The method is applied to the chassis domain controller, including: obtaining vehicle state data; determining the fifteen-degree-of-freedom dynamic model data of the vehicle according to the vehicle state data ; Obtain the first motion boundary, which is determined by the external intelligent driving domain controller based on the driving road data; determine the second dynamic boundary of the car according to the first motion boundary and the fifteen-degree-of-freedom dynamic model data; The second power boundary and the preset corresponding relationship determine the third execution boundary, so that the actuator on the vehicle controls the operation of the vehicle according to the vehicle control data within the third execution boundary. In this way, the current driving environment can be fully utilized under the condition of ensuring driving stability, and the disadvantage of adopting an overly conservative control strategy can be overcome. And it can easily and accurately obtain the execution amount boundary of the actuator, which alleviates the accuracy of the current control algorithm and the problem of computing resources.

The invention discloses an optical flow filtering method guided by a motion boundary based on a collaborative deep neural network, and constructs an optical flow filtering data set guided by a motion boundary and a collaborative deep neural network. The input is the initial optical flow estimation result and the motion boundary, and the output is the filtered Optical flow estimation results, including the initial optical flow feature extraction subnetwork, motion boundary feature extraction subnetwork, optical flow filtering first subnetwork and optical flow filtering second subnetwork, use the training set to train the collaborative deep neural network, and use the training to complete The collaborative deep neural network filters the initial optical flow estimation results to quickly generate higher-precision optical flow estimation results. The method of the present invention uses a cooperative deep neural network to automatically learn the optical flow filtering process guided by the motion boundary, accurately simulates the complex functional relationship from the input variable to the output optical flow filtering result, and avoids the introduction of new errors by irrelevant edge information other than the motion boundary , to improve the efficiency and accuracy of optical flow filtering.

A dual-mode muffler valve automatic pre-twisting welding detection equipment is composed of a bracket, a torque measurement system, a drive system, a limit system, a welding system and a control system. The torque measuring system, driving system and control system are respectively fixed on the bracket, and the limit system and welding system are installed on the driving system. The torque measurement system transmits the torque to the static torque sensor through the asbestos plastic joint to realize the measurement of the torque, and at the same time isolates the impact of the heat and current generated by welding on the sensor; the drive system realizes the automatic clamping of the valve through the combination of linear motion and rotary motion and pre-twist; the limit system controls the motion boundary position of the drive system through linear motion limit and rotary motion limit; the welding system realizes the automatic welding of the torsion spring seat of the valve through the combination of cylinder and slide plate; the control system controls each The orderly movement of the system makes the three processes of pre-twisting, welding and testing of the dual-mode muffler valve automatically completed.

The invention discloses a human behavior recognition method based on the fusion of multi-feature spatio-temporal relations. The specific steps include: expressing the dense track features extracted from the video through the optical flow histogram and the motion boundary histogram, and then using the KMEANS algorithm to construct the two features Corresponding to the spatio-temporal bipartite graph between the centroids, the K-way bipartite graph segmentation technology is used to segment the spatio-temporal bipartite graph, and the representation method based on conditional probability is used to obtain the video-level code after the fusion of the two features, and finally the classifier is trained and recognized , through the above method, the present invention is a human behavior recognition method based on multi-feature spatio-temporal relationship fusion, the method calculates the spatio-temporal distance between the features in each video, thereby constructing the spatio-temporal bipartite graph between the centroids corresponding to the two features , the K-way bipartite graph segmentation technology is used to segment the spatio-temporal bipartite graph, and the centroid with strong spatio-temporal relationship is fused, which better excavates the effective information of different features and improves the recognition accuracy.

The purpose of the present invention is to provide a multi-block structure grid division method for the unsteady flow field of a complex structure spring valve, which includes the following steps: extracting a flow field model from a design model of a spring check valve; The simplified two-dimensional watershed of the model is divided into moving and non-moving sub-basins; the height of the first layer grid is calculated, and each two-dimensional sub-basin is divided into structural grids; the upper and lower parts of the two-dimensional structured grid are generated into three-dimensional At the same time, the three-dimensional unstructured mesh of the middle part of the watershed is generated; the three-dimensional flow field meshes of the upper, middle and lower parts of the spring valve are integrated to obtain the overall flow field mesh of the valve group. The invention can provide a grid that conforms to the actual motion law of the model and facilitates the setting of dynamic grids for the numerical simulation of the dynamic flow field of the spring valve. The motion and static sub-regions in the watershed do not affect each other, and the block-structure grid is easy to generate dense motion. Boundary layer mesh, the accuracy of numerical simulation is greatly improved.