Device and method for detecting abnormal situation
AssigneeS-1 Corporation
InventorsJin Sun Ju, Hwijun Kim, Hyang Sook Choi
Life of this patent
Estimating the term…
Abstract
Provided is a detection method of a detection device. The detection device detects at least one object from an image by using depth information about the image obtained by a camera. The detection device identifies whether said at least one object is a person through three-dimensional (3D) head model matching, which matches a head candidate area of said at least one object with a 3D head model. The detection device calculates a feature for detection of a situation by using said at least one object when it is identified that said at least one object is the person.
Claims18
A detection method of a detection device, the detection method comprising:
- detecting at least one object in a frame by using depth information about the frame obtained by a camera;
- identifying whether said at least one object is a person through three-dimensional (3D) head model matching, which matches a head candidate area of said at least one object with a 3D head model;
- and
- calculating a feature for detection of a situation by using said at least one object when it is identified that said at least one object is the person, wherein calculating the feature for detection of the situation includes:
- tracking the at least one object in subsequent frames;
- and
- incrementing a number of movers in a first direction if a center point of said at least one subject passes a count line and a first virtual count line.
A detection device, comprising:
- an object detecting unit, which detects at least one object by using depth information about an image obtained by a camera;
- an identifying unit, which compares a head candidate area of said at least one object and a 3D head model and identifies whether said at least one object is an object of a target type;
- and
- a recognition feature calculating unit, which calculates a feature for detection of a situation by using said at least one object when it is identified that said at least one object is the object of the target type,
- wherein the identifying unit is configured to calculate at least one of a first pixel having a minimum depth value based on the camera among pixels of said at least one object and a second pixel having a maximum height value based on a floor area of the image, and designate a circle area, which is based on at least one of the first pixel and the second pixel having a maximum height value and has a predetermined radius, as a first head candidate area.
Dependencies are inferred from the claim text.