公开(公告)号：US20150227784A1

公开(公告)日：2015-08-13

申请号：US14614891

申请日：2015-02-05

Applicant: Tata Consultancy Services Limited

Inventor： Sangheeta ROY ,  Tanushyam CHATTOPADHYAY ,  Dipti Prasad MUKHERJEE

IPC: G06K9/00 ,  G06T7/00 ,  G06T7/40

CPC classification number: G06K9/00369 ,  G06K9/00201 ,  G06K9/00342 ,  G06K9/00771 ,  G06K9/40 ,  G06K9/4638 ,  G06K2009/485 ,  G06T2207/10024 ,  G06T2207/10028 ,  G06T2207/20021 ,  G06T2207/20072 ,  G06T2207/30196 ,  G06T2207/30232

Abstract: Disclosed is a system and method for detecting a human in an image, and a corresponding activity. The image is captured, wherein the image comprises a plurality of pixels having gray scale information and a depth information. The image is segmented into a plurality of segments based upon the depth information. A connected component analysis is performed on a segment in order to segregate the one or more objects into noisy objects and candidate objects, the noisy objects are eliminated from the segment. A plurality of features are extracted from the candidate objects, and are evaluated using a Hidden Markov Model (HMM) model in order to determine the candidate objects as one of the human or non-human. The corresponding activity associated with the human is detected based on a depth value associated with each pixel corresponding to the candidate object in the image.

Abstract translation: 公开了一种用于检测图像中的人的系统和方法以及相应的活动。 拍摄图像，其中图像包括具有灰度信息和深度信息的多个像素。 基于深度信息将图像分割成多个片段。 在段上执行连接分量分析，以便将一个或多个对象分离成噪声对象和候选对象，从该段中消除嘈杂对象。 从候选对象中提取多个特征，并且使用隐马尔可夫模型（HMM）模型来评估，以便将候选对象确定为人或非人之一。 基于与图像中的候选对象对应的每个像素相关联的深度值来检测与人相关联的相应活动。

公开(公告)号：US20190125238A1

公开(公告)日：2019-05-02

申请号：US16172668

申请日：2018-10-26

Applicant: Tata Consultancy Services Limited

Inventor： Sangheeta ROY ,  Oishee MAZUMDER ,  Debatri CHATTERJEE ,  Kingshuk CHAKRAVARTY ,  Aniruddha SINHA

IPC: A61B5/00 ,  A61B5/11 ,  G06N5/04 ,  G06F17/15 ,  G06T19/00

Abstract: Systems and methods for quantification of postural balance of users in an augmented reality (AR) environment. Traditional systems and methods provide for quantifying the postural balance using the AR environment but none of them quantify or restrict the functional tasks performed by the users to a predefined level. Embodiments of the present disclosure provide for the quantification of the postural balance with a variable step height in the AR environment by acquiring first set of information comprising of data on skeletal joints, filtering the first set of information for obtaining a filtered set of data, computing the set of postural data based upon the filtered set of data and quantifying the postural balance based upon the set of postural data by computing threshold values for obtaining postural stability index scores and determining based upon the postural stability index scores, the postural balance of the users in the AR environment.

公开(公告)号：US20160035124A1

公开(公告)日：2016-02-04

申请号：US14767161

申请日：2014-01-22

Applicant: TATA CONSULTANCY SERVICES LIMITED

Inventor： Aniruddha SINHA ,  Tanushyam CHATTOPADHYAY ,  Sangheeta ROY ,  Apurbaa MALLIK

IPC: G06T15/08 ,  G06T7/20 ,  G06T7/00

CPC classification number: G06T15/08 ,  G06T7/11 ,  G06T7/187 ,  G06T7/20 ,  G06T7/50 ,  G06T2207/10016 ,  G06T2207/10024 ,  G06T2207/10028 ,  G06T2207/30196 ,  G06T2207/30232

Abstract: Disclosed is a method for segmenting a plurality of objects from a two-dimensional (2D) video captured through a depth camera and an RGB/G camera. The method comprises detecting camera motion in each 2D frame of the plurality of 2D frames from the 2D video and generate a first set of 2D frames without any camera motion. The method further comprises generating a plurality of cloud points for the first set of 2D frames corresponding to each pixel associated a 2D frames in the first set of 2D frames. The method further comprises generating a 3D grid comprising a plurality of voxels. The method further comprises determining valid voxels and an invalid voxels in the 3D grid. Further, a 3D connected component labeling technique is applied on to the set of valid voxels to segment the plurality of objects in the 2D video.

Abstract translation: 公开了一种用于通过深度相机和RGB / G相机捕获的二维（2D）视频来分割多个对象的方法。 该方法包括从2D视频检测多个2D帧的每个2D帧中的相机运动，并且生成第一组2D帧而没有任何相机运动。 该方法还包括为与第一组2D帧中的2D帧相关联的每个像素对应的第一组2D帧生成多个浊点。 该方法还包括生成包括多个体素的3D网格。 该方法还包括确定3D网格中的有效体素和无效体素。 此外，3D连接的部件标注技术被应用于一组有效的体素以分割2D视频中的多个对象。