62 results about "Motion history" patented technology

Various techniques for temporally filtering raw image data acquired by an image sensor are provided. In one embodiment, a temporal filter determines a spatial location of a current pixel and identifies at least one collocated reference pixel from a previous frame. A motion delta value is determined based at least partially upon the current pixel and its collocated reference pixel. Next, an index is determined based upon the motion delta value and a motion history value corresponding to the spatial location of the current pixel, but from the previous frame. Using the index, a first filtering coefficient may be selected from a motion table. After selecting the first filtering coefficient, an attenuation factor may be selected from a luma table based upon the value of the current pixel, and a second filtering coefficient may subsequently be determined based upon the selected attenuation factor and the first filtering coefficient. The temporally filtered output value corresponding to the current pixel may then be calculated based upon the second filtering coefficient, the current pixel, and the collocated reference pixel.

A de-interlacer includes recursive motion history map generating circuitry operative to determine a motion value associated with one or more pixels in interlaced fields based on pixel intensity information from at least two neighboring same polarity fields. The recursive motion history map generating circuitry generates a motion history map containing recursively generated motion history values for use in de-interlacing interlaced fields wherein the recursively generated motion history values are based, at least in part, on a decay function.

The invention relates to a human behavior identification method based on the depth information. The method comprises steps that depth images are acquired by utilizing a depth camera; a two-dimensional depth image coordinate system is converted to a three-dimensional camera coordinate system; X, Y and Z values of a three-dimensional point under the three-dimensional camera coordinate system are assigned to X, Y and Z values of a point cloud three-dimensional point; multiple frame images of each motion which are converted into three-dimensional coordinates fill a three-dimensional body in a point cloud storage format, and a three-dimensional human motion history body is acquired; the three-dimensional human motion history body is divided into identification and training samples; a word frequency distribution histogram and a class statistics histogram of the training sample are acquired, and the acquired histograms are inputted to a SVM training determination motion classifier model; a word frequency distribution histogram of the identification sample is inputted to a classifier model to carry out identification operation, and thereby the identification result is acquired. According to the method, problems that depth value utilization is not enough and the specific space structure information is insufficient are solved through the generated three-dimensional human motion history body.

A method for object analysis using motion history is provided. The method includes receiving video data comprising a plurality of frames of a scene comprising one or more elements, and processing the video data to produce a motion history image comprising motion history values of at least one of the elements. The method also includes identifying the one of the elements for further processing if a characteristic of the motion history image satisfies a criteria.

A de-interlacer includes recursive motion history map generating circuitry operative to determine a motion value associated with one or more pixels in interlaced fields based on pixel intensity information from at least two neighboring same polarity fields. The recursive motion history map generating circuitry generates a motion history map containing recursively generated motion history values for use in de-interlacing interlaced fields wherein the recursively generated motion history values are based, at least in part, on a decay function.

The invention is applied to the field of video detection, and provides a video significance detection method. The method comprises the steps that superpixel segmentation is conducted on a to-be-detected current frame to obtain a current frame after superpixel segmentation, optical flow field motion estimation is calculated according the current frame and a previous frame, motion distribution energy and motion edge energy are calculated, motion history energy is calculated according to the current frame and the previous frame, and a hybrid motion energy diagram is generated according to the above-mentioned features and a significance diagram of the previous frame; and an initial target segmentation area of the hybrid motion energy diagram is obtained, a reliable target area and a reliable background area are extracted, and a significance global optimization model is constructed according to the reliable target area, the reliable background area and the hybrid motion energy diagram and then solved to obtain a significance diagram of the current frame. According to the video significance detection method, by adopting multiple motion features and space features such as the motion distribution energy of an area layer, the motion edge energy of an edge layer, the motion history energy of a pixel layer and the significance diagram of the previous frame, and the robustness and stability of significance detection are enhanced.

The invention relates to an abandoned object detection method based on computer vision, which mainly includes the following steps: updating motion history timing, and studying and updating a background model; by means of the background model, judging that each pixel is in one of three states: background, abandoned object foreground and non-abandoned object foreground, and carrying out state maintenance timing; extracting a connected region formed by the pixels constantly being in the same state of abandoned object foreground and timed at a preset value, and extracting objects to be analyzed; extracting characteristic expressions from the objects to be analyzed, and accumulating the number of the objects to be analyzed with the similar characteristic expressions within preset time; and if the accumulated value reaches a preset value, then carrying out comprehensive analysis on the motion characteristics of the corresponding objects to be analyzed in order to determine whether the objects to be analyzed are abandoned objects or not. The abandoned object detection method based on computer vision can quickly and accurately locate the corresponding regions of extracted abandoned objects, and has the characteristics of wide application range, high self-adaptability, high real-timeness, low missing rate and high reliability.

A real-time location system (RTLS) uses transmitting and listening tags, bridges, and beacons. The fixed beacons broadcast BLE advertisements containing motion-status information about recent history of perceived motion in. a room as determined from a motion sensor in the beacon. The bridges forward the beacon's received advertisements to a location engine, which records timestamps of motion events seen by each beacon in each room. One or more tags then report their own motion status based on a tag-based accelerometer. The system utilizes a series of location-engine steps, to estimate the room-location of the tags based on a specific combination of RSSI analysis, and a comparison of tag-motion history to the perceived arid recorded motion-status in a room. This analysis of tag-motion history and motion-in-room status produces a better estimate of room-level location of the tag than can be estimated by simple proximity or multi-lateration using radio signal strengths.

The invention discloses a precise strike-oriented ground moving target detection and recognition method. The method comprises the steps of: combining space-time information to extract a target candidate region, wherein the time-domain information is utilized to acquire moving images, forward-moving history graphs and backward-moving history graphs are calculated according to the moving images, images obtained by carrying out minimum-value acquisition on the forward-moving history graphs and the backward-moving history graphs according to pixels are used as input images, candidate moving regions are obtained through self-adaptive threshold processing and communication-domain extraction, the space-domain information is utilized to calculate materiality scores of the candidate moving regions , and the target candidate region is obtained through threshold processing; and recognizing the target candidate region, wherein feature learning is completed through local low constraint encoding, a linear support vector machine is utilized to realize recognition for the target candidate region, and a candidate region recognized as specific targets is reserved to obtain a final detection result. The method realizes automatic detection and recognition for the multiple ground moving targets in a complex environment, and improves the strike precision.

The invention relates to an RPC plate damage evaluation method based on a P-I curve under an action of explosion load. A explosion load point is evaluated according to explosive payload of TNT explosive and explosion distance; based on an equivalent unidirection method, motion histories of a certain point of an RPC plate component is evaluated, a midspan displacement time-history curve of the RPCplate under bending response is acquired, and the displacement peak value of the curve is acquired; an equivalent unidirection system under shearing response is established, a displacement time-history curve of a support seat position of the RPC plate under the shearing response is acquired, and the displacement peak value of the curve is acquired; a damage degree evaluation criterion is defined;the displacement peak value under the midspan bending response, the displacement peak value under the support seat shearing response and the damage degree evaluation criterion are utilized, and thus aP-I curve graph is drawn; and the explosion load point is drawn in the P-I curve graph, and the damage degree is determined. According to the RPC plate damage evaluation method based on the P-I curveunder the action of the explosion load, the possibility that shearing damages occur at the support seat position of the PRC plate component is considered, the evaluation method is safer, and comparedwith an existing evaluation method, the evaluation time can be shortened.

The invention discloses a human action recognition method based on a deep motion map (DMM) generated by a motion history point cloud (MHPC). The method comprises steps: the MHPC is generated; the DMMis generated; HOG feature vectors are extracted; an SVM classifier is trained and tested; and finally, the output of the SVM classifier is a human action classification result. The information of a human action at different view angles can be acquired, and the human action angle change robustness is enhanced. When the motion history point cloud projection is used to generate the deep motion map, coordinate normalization operation is carried out, and robustness on action intra-class differences is enhanced; and the HOG features extracted from the deep motion map generated by the motion historypoint cloud projection can effectively characterize human action classes, and the problem of complexity when the point cloud is used for feature extraction is solved.

A real-time location system (RTLS) uses Bluetooth Low Energy (BLE) transmitting tags, bridges, and beacons. The fixed beacons broadcast BLE advertisements containing motion-status information about recent history of perceived motion in a room as determined from a motion sensor in the beacon. The bridges forward the beacon's received advertisements to a location engine, which records timestamps of motion events seen by each beacon in each room. One or more simple transmitting tags then report their own motion status based on a tag-based accelerometer. The system utilizes a series of location-engine steps, to estimate the room-location of the tags based on a specific combination of RSSI analysis, and a comparison of tag-motion history to the perceived and recorded motion-status in a room. This analysis of tag-motion history and motion-in-room status produces a better estimate of room-level location of the tag than can be estimated by simple proximity or multi-lateration using radio signal strengths.

Methods of processing video are presented to generate signatures for motion segmented regions over two or more frames. Two frames are differenced using an adaptive threshold to generate a two-frame difference image. The adaptive threshold is based on a motion histogram analysis which may vary according to motion history data. Also, a count of pixels is determined in image regions of the motion adapted two-frame difference image which identifies when the count is not within a threshold range to modify the motion adaptive threshold. A motion history image is created from the two-frame difference image. The motion history image is segmented to generate one or more motion segmented regions and a descriptor and a signature are generated for a selected motion segmented region.

Disclosure is related to a system and a method for sport imagination. The system stores multiple players' sport historical archives. The data in the archives includes every player's sport path and moving positions with time. A user is authorized to obtain the sport historical archives, and an electronic map is applied to. When the sport historical archives are reproduced, the moving positions with time can be marked onto the electronic map. Therefore the user is allowed to get another player's instantaneous state when he runs at the same sport path. Simultaneously, the player's moving mileage, time, speed, and physiological information can be shown. Therefore, the system provides the user a virtual race circumstance.

Methods of processing video are presented to generate signatures for motion segmented regions over two or more frames. Two frames are differenced using an adaptive threshold to generate a two-frame difference image. The adaptive threshold is based on a motion histogram analysis which may vary according to motion history data. Also, a count of pixels is determined in image regions of the motion adapted two-frame difference image which identifies when the count is not within a threshold range to modify the motion adaptive threshold. A motion history image is created from the two-frame difference image. The motion history image is segmented to generate one or more motion segmented regions and a descriptor and a signature are generated for a selected motion segmented region.

The invention discloses a moving small target detection method under a complex ground background, and the method comprises the steps: extracting sparse optical flow points, and calculating a background motion estimation matrix; performing background motion compensation by using the motion estimation matrix to obtain a frame difference image; fusing multiple frames of frame differences before and after the frame difference graph to obtain a forward and backward motion history graph; performing threshold processing on the forward and backward motion history graph, and extracting a connected domain based on a region growing method to obtain a candidate motion target; carrying out data association on the multiple frames of candidate targets before and after to obtain a plurality of motion tracks; calculating a confidence coefficient score of each track according to the actual motion characteristics of the target; and according to the confidence coefficient of the track, removing and complementing the candidate targets to obtain a final small target detection result. According to the detection method provided by the invention, for the moving target detection problem of a complex background and a small-size target, candidate moving area extraction is carried out based on a historical graph, and multiple frames of moving information before and after fusion are fused, so that the comprehensiveness, accuracy and high precision of detection are guaranteed.

An electronic device is provided. The electronic device includes a processor configured to record exercise history information relating to an exercise performed by a user of the electronic device, and a memory configured to store the exercise history information, wherein the processor is configured to record the exercise history information continuously if an exercise application of the electronic device is unintentionally terminated, and retrieve the stored exercise history information when the exercise application is restarted.

The invention discloses a video emotion recognition method based on a local enhanced motion history map and a recursive convolutional neural network, and the method comprises the steps: 1, obtaining astatic expression image data set and an expression video data set, and carrying out the preprocessing of the expression video data set; 2, calculating a local enhanced motion history map LEMHI; 3, pre-training a convolutional neural network VGG16 model by using the static picture data set; 4, performing fine tuning on the pre-trained VGG16 model by using LEMHI to obtain a LEMHI-CNN model; 5, inputting the video frame into a pre-trained VGG16 model to extract spatial features; 6, stacking, fragmenting and pooling the spatial features, and training an LSTM neural network model to obtain a CNN-LSTM model; 7, performing weighted fusion on the identification result of the LEMHI-CNN model and the CNN-LSTM model to obtain a final identification result. According to the invention, the video emotion recognition rate can be obviously improved.

Described is a system for detecting moving objects using multi-frame motion history images. An input video sequence of consecutive registered image frames is received. The sequence of consecutive registered image frames comprises forward and backward registered image frames registered to a coordinate system of a reference image frame. Frame differences are computed between each of the consecutive registered image frames and the reference image frame. The frame differences are accumulated based on characteristics of the input video sequence to compute a corresponding motion response value. A selected threshold value is then applied to the motion response value to produce at least one binary image used for detection of moving objects in the input video sequence. Additionally, the invention includes a system for adaptive parameter optimization by input image characterization, wherein parameters that are based on characteristics of the image influence the motion detection process.

The invention discloses a video target detection method based on a motion history image, and aims to improve the speed and accuracy of video target detection. The video target detection method comprises three aspects: (1) for an input video frame sequence, calculating motion history images of the video frame sequence, and performing feature extraction on video frames and the motion history imagesthereof through a residual network; (2) fusing the extracted two parts of features, and inputting the fused features into a convolutional neural network to extract candidate boxes; and (3) obtaining avideo target detection result according to a bounding box regression algorithm and the constructed classifier. According to the video target detection method, the motion history images are added intothe model training process, so that the feature information of the video frames is provided for the model, and the association information between the video frame sequences is increased, and the accuracy of video target detection can be improved.

The invention discloses a human body behavior recognition method based on a spatio-temporal distribution map generated by motion history point clouds. The human body behavior recognition method comprises the following steps: generating an MHPC (Motion History Point Cloud); generating an STDM (Spatio-Temporal Distribution Map); extracting a color moment feature vector; extracting an LBP feature vector; training and testing a KELM classifier, and finally fusing output results by adopting a decision layers to obtain a human body action type label. The human body behavior recognition method disclosed by the invention can acquire information of human body actions under different visual angles, so that the robustness of an action angle change is improved. The STDM for expressing a human body action is more comprehensive than a depth image, and extracted features are more distinctive; extracted color moment feature and LBP feature can effectively characterize human body action types, so thatthe problem of complexity in feature extraction by using the point clouds is solved. By use of decision layer-based fusion for classification, the shortcomings of incompatibility and high dimension offeature layer fusion can be avoided.

The invention relates to a group abnormal behavior identification method, which is used for carrying out series pedestrian detection on the basis of an acquired video stream, calculating an energy value E on the basis of a motion history graph when multiple persons are judged, judging that multiple persons have abnormal behaviors when E is greater than or equal to a set energy threshold, and giving an alarm. From the overall perspective, significant features for distinguishing abnormal behaviors and normal behaviors are found for analysis and judgment, and the normal behaviors and the abnormalbehaviors of the group are distinguished in combination with crowd density estimation and the energy value of the motion history graph, and the accuracy of the recognition algorithm is high, and thereal-time requirement of the system is met.

The present application provides a method of evaluating spindle component wear in a turbine by a data acquisition system. The method may include the steps of receiving a number of operating parameters from a number of sensors, wherein the operating parameters may include spindle position from which a number of spindle strokes and an accumulated spindle distance may be determined and temperature, generating a movement history based on the operating parameters, comparing the movement history to predetermined values, and altering one or more of the operating parameters and / or initiating repair procedures if the movement history exceeds predetermined values.

The invention discloses a method for detecting abnormal behaviors in an elevator car. The method comprises: carrying out preprocessing operation on an original image of an elevator monitoring video; screening the preprocessed videos, and segmenting video segments in which people, pets and electric vehicles appear by adopting an optical flow method; inputting the screened video segments into a trained yolov3 model for recognition to obtain the number of people and the number of pets in the video, and judging whether an electric vehicle appears or not; calculating a motion history map of each frame in the screened video segment, and calculating an energy value of each frame image according to the motion history map; adaptively determining an energy threshold according to the number of peopleand the number of pets; and combining the energy value of the image with the determined adaptive energy threshold, and determining whether the electric vehicle exists, and judging whether an abnormalbehavior occurs in the video segment. And the video segments with the targets are screened out and then processed, so that the computing power consumption of the servers is reduced, the requirementson the number and configuration of the servers are reduced, and the operation cost is also reduced.

The invention provides a human behavior identification method based on a space-time distribution map generated based on motion history point cloud. The method comprises the following steps: generatingthe MHPC; generating the STDM; extracting HOG feature vector; and training and testing a KELM classifier, and finally, the output of the KELM classifier being a human body motion classification result and the like. The method can obtain information of body motion from different angles of view, thereby enhancing robustness of motion angle change; body motion information is more complete when represented by the space-time distribution map than represented by a depth motion map, and the extracted features are also more distinctive; the extracted HOG features can effectively represent human action categories, and thus the problem of complex feature extraction based on the point cloud is solved; and by adopting an extreme learning machine based on a Gaussian kernel function, the method has theadvantages of high classification precision and fast learning speed.

The invention discloses a human body behavior recognition method based on a history motion graph and R transformation. According to the method, a depth video is used as a recognition basis, firstly, the minimum enclosure rectangle of human body motion is calculated according to a foreground segmentation technology, then the history motion graph is extracted within a depth video area limited by the minimum enclosure rectangle, motion intensity constraint is exerted on the extracted history motion graph, so that a motion energy diagram is obtained, R transformation is calculated on the obtained motion energy graph, and therefore a characteristic vector used for behavior recognition is obtained. A method of a support vector machine is adopted for training and recognition processes. The minimum enclosure rectangle of human body behavior motion is adopted for preprocessing, and behavior characteristic extraction is accelerated; a method of history motion graph sequences is adopted for reducing influences of noise in depth graphs; characteristics are extracted through performing R transformation on the energy graph, so that calculation speed is high.

The invention discloses a detection and tracking method suitable for use on a ship. The method comprises the following steps: (1) using a Canny operator and combining water-front boundary contour information to extract a water-front boundary line in a case of no ship from an input video sequence; (2) extracting a motion history graph; (3) carrying out morphological processing on a water surface region in the motion history graph; and (4) reconstructing ship contour circumscribed-rectangles. According to the method, a river-front region interfering with detection is separated through extractingthe water-front boundary line, impacts of moving targets on a shore are eliminated, morphological processing is rationally applied to eliminating interference of non-moving targets on a water surface, the contour circumscribed-rectangles are combined to reconstruct and complement a motion trajectory with deficiency due to too low speed, and accurate detection and real-time tracking of the ship under a complex dock environment are realized.

The invention discloses an intelligent wardrobe. The intelligent wardrobe comprises a wardrobe body, an identification unit, a transmission mechanism and a controller, and the controller comprises aninformation receiving unit, a first processing unit, a second processing unit and a third processing unit; the information receiving unit is used for receiving route and weather information on a mobile phone, movement records and average heart rate data of a business traveler; the first processing unit is used for selecting proper clothes in the wardrobe body according to the route information, the second processing unit judges the emotion of the business traveler for a current travel route according to the average heart rate data, further screening is conducted on the basis, and the third processing unit is used for correcting a screening result of the second processing unit according to historic movement records. The proper clothes can be accurately selected for the business traveler according to the travel route and the movement records and the emotion of the business traveler, the time is saved for the business traveler, the mood of the business traveler is improved, and the intelligent life experience is provided for the business traveler.

In order to maintain the space-time coherence of a redirected video and the reconstruction quality of important contents of the video, the invention provides an intelligent cropping video redirectionmethod based on relative displacement constraints. The method comprises the following steps: firstly, adaptively fusing a spatial saliency map and a motion history map to obtain an importance map of each frame; obtaining relative displacement between frames through an SIFT matching algorithm and an RANSAC method, and adding the relative displacement between the frames into a time coherence constraint term; carrying out scaling optimization on the video groups after scene detection grouping, and determining an optimal redirection window and a scaling factor of the video groups; and finally, redirecting the video group by using the corresponding redirection window and scaling factor. And if the jump frame exists, performing smoothing processing on the importance degree graph through a feedback mechanism, and performing scaling optimization and redirection again to reduce jitter. Experimental results show that compared with an existing video redirection method, the method not only can obtain better reconstruction quality of important content of the video, but also effectively reduces jitter of the redirection video.