Simultaneous metadata extraction of moving objects

摘要

In some examples, a main computing device may be configured to segment the video stream into multiple video clips processing a video stream that includes multiple frames. Further, the main computing device may be configured to distribute the video clips to multiple computing nodes to extract the metadata of moving objects in the video clips. The extracted metadata of a same moving object that appears in different video clips may be further merged together by the main computing device.

主权项

1. A method to simultaneously process multiple portions of a video stream, comprising:
segmenting, by a computing device, the video stream that includes multiple frames into multiple video clips; calculating, by the computing device, a computational complexity value for each of the multiple video clips; receiving, by the computing device, computational capability information regarding each of multiple computing nodes; respectively distributing, by the computing device, the multiple video clips to the multiple computing nodes in accordance with the received computational capability information of each of the multiple computing nodes and the respective computational complexity value for each of the multiple video clips to allow for simultaneous extraction of metadata of moving objects in the video stream on the multiple computing nodes, each of the multiple computing nodes being configured to:
identify one or more moving objects in each of the multiple video clips, andextract metadata corresponding to each of the one or more moving objects from each of the multiple video clips; receiving, by the computing device, the extracted metadata from each of the multiple computing nodes; and with respect to one of the multiple video clips, combining, by the computing device, the extracted metadata of one of the one or more moving objects with the extracted metadata of the one of the one or more moving objects in another one of the multiple video clips when the one of the one or more moving objects appears in the another one of the multiple video clips that immediately precedes or follows the one of the multiple video clips.