System for finding archived objects in video data

• US 8,160,371 B2
• Filed: 12/03/2007
• Issued: 04/17/2012
• Est. Priority Date: 12/03/2007
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Ma, Yunqian, Miller, Ben A., Johnson, Andrew H., Markham, Thomas R.
• Primary Examiner(s): Allison, Andrae S
• Application Number:
  US11/949,519
• Publication Number:
  US 20090141993A1
• Time in Patent Office: 1,597 Days
• Field of Search: 382/103, 382/107, 382/108, 382/173, 382/206, 382/232, 382/276, 382/293, 382/305, 382/236, 345/16, 345/3.2, 345/721, 345/719, 345/328, 345/432, 345/427, 345/430, 348/155, 348/335, 348/584, 348/700, 375/240.16
• US Class Current: 382/226
• CPC Class Codes: G06F 16/532 Query formulation, e.g. graphical querying
• CPC Class Codes: G06F 16/70 of video data
• CPC Class Codes: G06V 10/7557 based on appearance, e.g. active appearance models [AAM]

Abstract

A system includes a motion detection processor, a motion tracking processor, a people detection processor, a controller, a fusion processor, an appearance model generator processor, a database, a fast search processor, and a matching processor. The motion detection processor, the motion tracking processor, the controller, the people detection processor, the fusion processor, and the appearance model generator processor comprise an analytics pipeline, and the database and the fast search processor comprise a data index pipeline.

Claims

1. A non-transitory computer readable medium comprising instructions that when executed by a processor execute a process comprising:

receiving an image frame from a video sensing device;

detecting motion in the image frame;

tracking the motion in the image frame;

generating an appearance model for an object detected in the image frame;

storing the object appearance model in a database;

building a tree-based structure including a fast search index, wherein the tree-based structure comprises one or more of a balanced tree and an unbalanced tree, wherein the balanced tree is based on a median distance between features of the object appearance model, and the unbalanced tree is based on clustering of features of the object appearance model;

requesting the image frame and detecting a person in the image frame;

generating an appearance model for a person detected in the image frame; and

searching a queried person'"'"'s appearance model in the image frame using the tree-based structure;

wherein detecting the motion and tracking the motion operate on a number of image frames in a video sequence, and detecting a person operates on a subset of the number of image frames in the video sequence.

2. The non-transitory computer readable medium of claim 1, wherein the object appearance model and the person appearance model are indexed in the tree-based structure.

3. The non-transitory computer readable medium of claim 1, comprising instructions for:

selecting a person in a current image frame;

querying the tree-based structure a plurality of times to locate appearance models similar to the selected person;

analyzing the located appearance models to determine whether the selected person matches one or more appearance models in the database; and

displaying a result on a user interface.