67 results about "Motion flow" patented technology

Techniques are disclosed for detecting foreground objects in a scene captured by a surveillance system and tracking the detected foreground objects from frame to frame in real time. A motion flow field is used to validate foreground objects(s) that are extracted from the background model of a scene. Spurious foreground objects are filtered before the detected foreground objects are provided to the tracking stage. The motion flow field is also used by the tracking stage to improve the performance of the tracking as needed for real time surveillance applications.

A method for determining an ego-motion of a vehicle is described, which is carried out in driver assistance systems, particularly in parking assistance systems. The method involves: taking a sequence of images, successive in time of a vehicle surroundings by a vehicle camera; determining, based on the image sequence, at least one motion flow with regard to an object in the vehicle surroundings; and determining the ego-motion of the vehicle based on the at least one motion flow.

Techniques are disclosed for detecting foreground objects in a scene captured by a surveillance system and tracking the detected foreground objects from frame to frame in real time. A motion flow field is used to validate foreground objects(s) that are extracted from the background model of a scene. Spurious foreground objects are filtered before the foreground objects are provided to the tracking stage. The motion flow field is also used by the tracking stage to improve the performance of the tracking as needed for real time surveillance applications.

The present invention discloses a video frame rate up-conversion method and system for improving motion fluency intelligently. The method comprises a first step of pre-analyzing two frames of continuous original video image; a second step of performing motion estimation on the two frames of image; a third step of calculating a new interpolated frame between the two frames of original image according to estimated motion information; a fourth step of performing post processing like hole filling, block-effect filling, image quality evaluation on the interpolated frame image; and a fifth step of according to information of the four previous steps, determining whether to replace the interpolated frame with a repeated frame. On the basis of the method, the present invention provides a video frame rate up-conversion system for improving motion fluency intelligently. Due to the frame rate up-conversion, stream of each processing module is controlled, so that high reuse and high parallelism of the information of all modules are realized. Through adoption of the method and system, high quality video frame rate improvement is realized at a source end, and the video motion fluency is improved intelligently.

The invention discloses self-adaptively selecting global motion estimation method for panoramic video coding, which comprises the following steps that: (1) an image is divided into a plurality of sub-image blocks, (2) zero vector block is decided, (3) multi-resolution motion is estimated, (4) translational motion model motion parameter is estimated, (5) six-parameter motion model motion parameter is estimated, (6) twelve-parameter motion model motion parameter is estimated, and (7) RDO optimizes an estimation model of self adaptively choosing global motion. The advantage is that the estimation model of self adaptively choosing global motion optimized by the RDO can well describe the camera motion flow characteristic of the current frame well so that the motion estimation is more accurate and the coding performance is higher. At the same time, aiming at the characteristics of the rich detail and the high resolution ratio of panoramic video motion, the invention takes the Lagrange parameter Lambada=0.85x2QP / 3+2QP / 6 and further obtains a good balance between compression code rate and image distortion degree.

The invention provides a stop-motion animation shooting control device and a control method thereof. The device comprises a moveable pedestal, an X-axis axial moving mechanism, a Y-axis axial moving mechanism, a Z-axis axial moving mechanism, a horizontal transverse beam, a holder and an electrical control mechanism, wherein photoelectric detectors are arranged at the beginning and end positions of each axis axial moving mechanism; and the holder is used for fixing shooting equipment. The device is characterized in that a multidirection rotating mechanism is arranged between the horizontal transverse beam and the holder. The invention also discloses the control method capable of adjusting the holder so as to adapt to different shooting angles. The device has the beneficial effects that the motion curve of the holder can be set in advance, the motion quantity of the control device is calculated through controlling a data calculation module and the shooting points are distributed uniformly, so that the camera lens can move smoothly and naturally during shooting.

A system and method for modeling articles of clothing includes a database having representation of an article of clothing and associated movement parameters characterizing motion flow of the article of clothing on a person in motion, and an image generator configured to generate a representative image of a person in motion. The movement parameters for the article of clothing are used to manipulate a representation of the article of clothing on a representation of the person in motion using the movement parameters for the article of clothing, thereby simulating the person in motion wearing the article of clothing.

The invention discloses a multi-motion stream depth convolution network model method for video prediction, which comprises the following steps: constructing a new convolution automatic coding networkframework fusing long-term and short-term memory network modules; the motion stream is proposed as the motion conversion from input frame to output frame. The method of generating multiple motion streams at the same time to learn more exquisite motion information is adopted to improve the prediction effect effectively. The base image is proposed as a pixel-level supplement to the motion flow method, which improves the robustness of the model and the overall effect of the prediction. A bilinear interpolation method is used to interact with multiple motion streams on the input frame to obtain multiple motion prediction maps, and then the motion prediction maps are linearly combined with the base image according to the weight matrix to obtain the final prediction results. By adopting the technical proposal of the invention, the time information in the video sequence can be more fully extracted and transmitted, thereby realizing longer-term, clearer and more accurate video prediction.

The present invention discloses an image super-resolution reconstruction method and a system. According to the method, firstly, the pixel data of a plurality of low-resolution images having the complementary information are acquired. Secondly, according to the pixel data, data consistency items selectively integrating brightness constancy constraints with gradient constraints are constructed and matched feature points are subjected to the initial flow vector updating treatment. Initial flow vectors are substituted into a motion estimation variation model constructed based on data consistency, and then the optimal motion flow vector of each pixel is obtained. Simultaneously, an optimal fuzzy core is constructed based on the brightness of the L0 norm and the fuzzy estimation energy function of a gradient-combined constraint image prior model. Finally, according to the optimal motion flow vector and the optimal fuzzy core, a reconstructed high-resolution image of a variational model is established. Based on the method or the system, the accuracy of motion displacement vectors and the accuracy of fuzzy parameters during the super-resolution reconstruction process can be effectively improved. As a result, an optimal super-resolution reconstruction result is obtained.

A method and system of real-time obstacle detection from a moving vehicle is provided. The method and system use a calibrated image capturing device. The method and system use a motion estimation technique to pick points with reliable image motion flows, and performs very fast sparse matching between the image motion flows and true motion flows calculated from the ego-motion of the image capturing device. Any mismatch between the image motion flows and the true motion flows are verified over time to achieve robust obstacle detection.

The invention provides a method for recognizing actions on the basis of deep feature extraction asynchronous fusion networks. The method is implemented by the aid of main contents including coarse-grained-to-fine-grained networks, asynchronous fusion networks and the deep feature extraction asynchronous fusion networks. The method includes procedures of inputting each short-term light stream stackof each space frame and each movement stream of input video appearance stream into the coarse-grained-to-fine-grained networks; integrating depth features of a plurality of action class grain sizes;creating accurate feature representation; inputting extracted features into the asynchronous fusion networks with different integrated time point information stream features; acquiring each action class prediction results; combining the different action prediction results with one another by the deep feature extraction asynchronous fusion networks; determining ultimate action class labels of inputvideo. The method has the advantages that deep-layer features can be extracted from the multiple action class grain sizes and can be integrated, accurate action representation can be obtained, complementary information in a plurality of pieces of information stream can be effectively utilized by means of asynchronous fusion, and the action recognition accuracy can be improved.

The invention discloses a ports interactive animation producing method. In the method, a layered frame is adopted to built a model for the interactive motion flow, namely in a state layer and in a motion layer. In the state layer, a finite state machine is built to simulate the conversion between the states in the interactive motion. In the motion layer, a mixed method is used to produce the motion choice according with the opposing situation between the present parts in an opposing state. For solving the problem that the data is captured by the single-part motion without the interactive action between the parts, the invention adopts a machine study and experience rule mixed method to capture and simulate the interactive action between the parts, the good effect is obtained, the produced animation has visual nature and trueness.

The invention discloses a behavior recognition method based on a space-time attention enhancement feature fusion network, and belongs to the field of machine vision. According to the method, a networkarchitecture based on an appearance flow and motion flow double-flow network is adopted, and is called as a space-time attention enhancement feature fusion network. Aiming at a traditional double-flow network, simple feature or score fusion is adopted for different branches, an attention-enhanced multi-layer feature fusion flow is constructed to serve as a third branch to supplement a double-flowstructure. Meanwhile, aiming at the problem that the traditional deep network neglects modeling of the channel characteristics and cannot fully utilize the mutual relation between the channels, the channel attention modules of different levels are introduced to establish the mutual relation between the channels to enhance the expression capability of the channel characteristics. In addition, thetime sequence information plays an important role in segmentation fusion, and the representativeness of important time sequence features is enhanced by performing time sequence modeling on the frame sequence. Finally, the classification scores of different branches are subjected to weighted fusion.

The invention relates to a novel method and equipment for simultaneously rolling and polishing the inner surface and the outer surface of a large thin-wall cylindrical workpiece. The method for rolling and polishing integrates a main shaft mode, a vibration mode and a rotation mode. Due to the loading and unloading mode that a linear motion unit drives a support to lift the workpiece, the mountingmode that a spindle box fixes the workpiece, the outer wall machining mode that the outer surface of the workpiece is covered with a grinding block and grinding fluid, the inner wall machining mode that the workpiece rotates for rolling and polishing, and the mode of generating a motion flow field under driving of a vibration motor, the automatic equipment can simultaneously roll and polish the inner wall and the outer wall of the large thin-wall cylindrical workpiece.

An arm sleeve with a hollow body made of a mesh material. Both ends of the body have an elastic band, one elastic band being smaller than the other. A fastener, such as one having engaging hooks and loops, is at the larger of the two elastic bands. Strips, flags or ribbons dangle from the underside of the body and are slightly spaced apart from each other in succession.

A waist-mounted parachute deployment and retracting system for training runners and other athletes that provides a user a method for selectively deploying and retracting the parachute during training without stopping or breaking stride. The parachute provides drag to increase resistance to improve performance by overloading the muscles. The user selectively deploys the parachute by pressing a button to release and open the parachute. Air flow created by the user's forward motion flows through a pair of ports on the side of a case holding the parachute. The air flow expands the parachute, causing the cords attaching the parachute to the case to extend and wind a recoil mechanism. The user selectively retracts the parachute by pressing a button to close the parachute and the recoil mechanism rewinds the cords, pulling the parachute into the case, ready to be deployed again when desired.

The present invention comprises using error propagation for building feature spaces with variable uncertainty and using variable-bandwidth mean shift for the analysis of such spaces, to provide peak detection and space partitioning. The invention applies these techniques to construct and analyze Hough spaces for line and geometrical shape detection, as well as to detect objects that are represented by peaks in the Hough space. This invention can be further used for background modeling by taking into account the uncertainty of the transformed image color and uncertainty of the motion flow. Furthermore, the invention can be used to segment video data in invariant spaces, by propagating the uncertainty from the original space and using the variable-bandwidth mean shift to detect peaks. The invention can be used in a variety of applications such as medical, surveillance, monitoring, automotive, augmented reality, and inspection.

The invention discloses a video space-time feature extraction method in the technical field of computer video processing. The method comprises the following steps of: performing Gaussian filtering on two adjacent frames of images in a video flow, calculating geometric regularity and geometric regularity direction of an edge and a texture region on the smooth image and constructing a spatial geometric flow field; establishing a motion equation between the two frames of images and calculating a motion vector of each pixel point by adopting a block matching correlation method; eliminating singular values of the calculated geometric flow field by using an M*M neighborhood local average method; and finally merging the spatial geometric flow field with the time domain geometric motion flow to obtain a space-time feature vector of the pixel point. The method has higher robustness and calculation performance for illumination change; a moving target has a clear structure; close moving targets can be effectively separated; and the method has higher robustness and efficiency for traffic flow estimation and abnormal event detection in traffic surveillance.

The invention provides a method and a system for simulating operation of a storage power station, a medium and an apparatus. Firstly, three-dimensional modeling is carried out for the hydraulic junction structure, equipment and operation tools of the storage power station. However, the movement mode, the state and the effect after the movement of the equipment model and the operation tool model are made; next, a finite state machine is constructed for each typical operation order. When the simulation of a typical operation order is to be carried out, the finite state machine of the typical operation order is started, each state in the finite state machine is triggered by the input condition to enter the running state, and the control equipment model and / or the operation tool model completethe corresponding motion flow according to a certain motion mode. The invention can effectively improve the effect of the operation and maintenance training of the storage power station, reduce the operation risk of the operation and maintenance on-the-spot practical operation training and reduce the manufacturing cost of the operation courseware of the storage power station according to the actual demand of the storage power station through the finite state machine.

The invention discloses a face image sight correction method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps of obtaining a to-be-corrected eye image from a face image; generating an eye motion flow field and an eye contour mask based on the to-be-corrected eye image; on the basis of the eye movement flow field and the eye contour mask, carrying out sight correction processing on the to-be-corrected eye image to generate a corrected eye image; and generating a human face image after sight correction based on the corrected eye image. The corrected eye image is directly generated end to end according to the input to-be-corrected eye image, the head posture angle does not need to be known, the time consumption is greatly reduced, the correction accuracy is improved, the sight correction function truly meets the actual available standard, and the method and device are suitable for video conferences, video calls, video live broadcast and other scenes.

The present invention comprises using error propagation for building feature spaces with variable uncertainty and using variable-bandwidth mean shift for the analysis of such spaces, to provide peak detection and space partitioning. The invention applies these techniques to construct and analyze Hough spaces for line and geometrical shape detection, as well as to detect objects that are represented by peaks in the Hough space. This invention can be further used for background modeling by taking into account the uncertainty of the transformed image color and uncertainty of the motion flow. Furthermore, the invention can be used to segment video data in invariant spaces, by propagating the uncertainty from the original space and using the variable-bandwidth mean shift to detect peaks. The invention can be used in a variety of applications such as medical, surveillance, monitoring, automotive, augmented reality, and inspection.

The invention provides a crowd movement flow drawing method based on sparse trajectory data. The method comprises the following steps: step 1, extracting a sparse track by using a movement extractionmethod, performing sorting according to time, connecting records before and after the current position of a track sequence, taking every two adjacent records as a group, and taking the space coordinates of the previous record as a starting point and the space coordinates of the later record as an end point in each group to obtain a local movement vector; step 2, performing space-time division on the motion vector in the step 1; 3, smoothing the motion vector in the step 2 by using a kernel density estimation method, and then performing aggregating to obtain a motion vector cluster in a local space; step 4, processing the motion vector in the step 2; step 5, clustering the motion vectors in the step 3 to obtain a significant global motion flow; and step 6, circularly displaying the global movement flow in the step 5, and displaying the abnormal value and the crowd distribution attribute obtained in the step 4.

The invention discloses a face image sight correction method and device, equipment and a storage medium, and relates to the technical field of artificial intelligence. The method comprises the steps of obtaining a to-be-corrected eye image from a face image; generating an eye movement flow field based on the to-be-corrected eye image and the target sight line direction; performing sight correction processing on the eye image to be corrected by adopting the eye movement flow field to generate a corrected eye image; and generating a human face image after sight correction based on the corrected eye image. According to the technical scheme provided by the invention, sight correction can be carried out without fixing the head posture, so that the technical scheme provided by the invention has relatively good sight correction capability for scenes such as video conferences, video calls, live videos and the like in which the head posture of the user changes in real time.

The invention relates to a ground robot positioning method based on multiple sensors. The method comprises the following steps that an encoder, a camera and an IMU are mounted on a robot; a plurality of two-dimensional codes are laid on the ground as map marks; in the advancing process of the robot, an encoder odometer module fuses data of an encoder and an IMU through a Kalman filter to provide continuous inertial track plotting, and a vision-inertial navigation positioning module fuses an image, the IMU, an encoder odometer track and motion manifold constraint through a nonlinear optimization algorithm to solve the six-dimensional pose of the robot; and the two-dimensional code detection module provides global pose correction. According to the method, various data information such as environment textures, two-dimensional codes, inertia measurement and encoders are efficiently fused, different fusion strategies are adopted for different data types, and high-precision robot six-dimensional pose positioning is provided.