8 results about "Motion history" patented technology

The application discloses a motion classification method for a dynamic vision sensor and relates to the field of motion classification. The application first adopts an event index plane, considers the space-time information in the cross neighborhood of each event in an event stream, calculates the motion gradient direction of each event as the local motion feature of each event in the event stream, then does pooling and pulse coding in the space-time dimension, and finally inputs the pulse sequence into a pulse neural network to output the classification result of the motion in the scene. The application introduces the event index plane with speed invariance to record the global motion history information of the motion, introduces the motion history information and the gradient direction calculation into the processing of the event stream, and proposes a novel motion gradient direction feature expression method based on events, which can efficiently extract the motion information of the motion and improve the motion recognition accuracy. In addition, compared with the optical flow motion feature expression, the proposed feature expression method avoids the complicated optical flow estimation process and saves the calculation energy consumption.

The application provides a motion history-based foreground background detection method, comprising the following steps: S1, preprocessing: denoising and smoothing, and enhancing the accuracy of frame difference; S2, calculating absolute frame difference; S3, binary processing: diff = absdiff (frame_blurred, prev_frame_blurred); and S4, updating a motion history image MHI. The method can avoid the adverse effects in the prior art, is simple to test, has high testing accuracy, is completely independent, is not interfered by a system, has very small error, and can be completely realized by a software module and is simple to operate.

The invention provides a concrete slump detection method and system for a discharge port of a concrete mixer, and the method comprises the steps: obtaining a video image sequence of a discharge port region, extracting spatial-temporal feature points from continuous video frames, carrying out the matching of a feature descriptor fusing local binary pattern texture and motion historical information, and generating an initial displacement vector field; local decomposition is carried out on the initial displacement vector field, abnormal matching points are eliminated according to a hydrodynamic incompatibility threshold value of velocity divergence and rotation, and a sparse matching flow field is obtained; constructing a geodesic line active contour model guided by the sparse matching flow field, and iteratively segmenting a concrete edge contour; taking the edge contour as a boundary, taking a velocity vector of the sparse matching flow field as a surface gradient, adopting a Poisson reconstruction algorithm to construct a three-dimensional surface model of the concrete accumulation body, and based on the projection area, the centroid height and the volume of the three-dimensional surface model, and combining a pre-calibrated mapping relation between geometric characteristics and slump, calculating the slump of the concrete accumulation body. And determining the concrete slump value.

This technology predicts the likelihood of slowing down during a marathon and provides users with useful information. [Solution] In the running analysis system 100, a prediction model is stored in the running analysis server 60 that predicts the pace changes in the latter half of the race using machine learning with motion history data that is measured over time from the beginning to the end of the race, using features corresponding to multiple types of motion analysis indicators, including at least running pace, as features that indicate the runner's motion state in a completed marathon. The measurement value acquisition unit 40 acquires motion history data that is measured over time with features corresponding to multiple types of motion analysis indicators during runs of less than the marathon race distance by the user 10, who is a marathon runner. Based on the motion history data measured over time during the user 10's run and the prediction model, the running analysis server 60 predicts the pace changes in the latter half of the race for the user 10 and notifies the user 10 of the prediction result.

A technique is directed to methods and systems for assessing wear for a ground engaging tool (GET), such as a tooth on a bucket on a machine. A GET monitoring system can utilize convolutional neural networks for detecting individual instances of GETs in an image. The GET monitoring system can records the initial state of a GET on a bucket during a calibration process. After calibration, the GET monitoring system can estimate how much the GET has worn down, as length or surface area, since the GET was installed. In some embodiments, as a machine operates the motion of the machine can induce perspective distortion on the captured images of the GETs. The GET monitoring system can correct for perspective distortion, induced by bucket motion, by building a motion history map of the GET.

The present application provides a server operation method for monitoring and mutual evaluation according to motion matching, a stage of obtaining initial information of a first student about a target sports project through a student terminal of the first student; a stage of creating at least one group for a coach of a plurality of students including the first student according to the initial information; a stage of receiving information of a student of a second group of students of the first student from a student of a first group of students matched with a first group of lecturers including the first student according to the information of the first group of students and the second group of students of the user, and training according to the information of the first group of students and the second group of students of the user. In the generation case, a stage of providing the first student motion history information and motion evaluation information as a lecturer terminal of a second lecturer matched with the second group can be included.

Provided is an information processing method for generating data using a target amount acquired by a moving body and position information and orientation information of the moving body. The data is generated by: identifying a plurality of recording target positions in the lateral width direction of the moving body on the basis of the position information and the orientation information of the moving body; and calculating post-division target amounts, each obtained by apportioning the target amount to the recording target positions at a predetermined ratio, and associating the post-division target amounts with the recording target positions. Thus, in order to more appropriately evaluate quantitative data acquired by the moving body, the method divides the data into a plurality of positions in consideration of the evaluation of the movement history, and associates the divided data with the position information.

The invention relates to the technical field of engineering safety guarantee, and particularly discloses a pipe gallery remnant real-time detection method and system, and the method comprises the steps: obtaining a current frame and a background frame of a picture at a current moment, and judging whether the current frame of the picture is a motion frame or not through combining with a preset object static displacement threshold value; if the current frame of the picture is a motion frame, a foreground object recognition image is obtained through a foreground object recognition network, and time sequence behavior analysis of a foreground object is further introduced to achieve remnant object judgment. Therefore, the lightweight judgment of the motion frame is that a large number of unchanged frames are filtered out from a subsequent model, so that the real-time detection of the remnant on a common server or edge computing equipment becomes possible, the time sequence behavior analysis of a foreground object is introduced, and the motion history of the object is taken as a criterion; according to the method, the crossing from object identification to object motion state understanding is realized, so that illegal missing objects are accurately identified, the false alarm rate of monitoring is reduced, and it is ensured that the output final result is highly credible.