Method and apparatus for detecting suspicious activity using video analysis
View Patent ↗A system detects a transaction outcome by obtaining video data associated with a transaction area and analyzing the video data to obtain at least one video transaction parameter concerning transactions associated with the transaction area. The transaction area can be a video count of items indicated in the video data as detected by an automated item detection algorithm applied to the video data. The system obtains at least one expected transaction parameter concerning an expected transaction that occurs in the transaction area, such as a scan count of items scanned at a point of sale terminal. The system automatically compares the video transaction parameter(s) to the expected transaction parameter(s) to identify a transaction outcome that may indicate fraudulent activity such as sweethearting in a retail environment.
1. A method of detecting suspicious activity, the method comprising:
receiving real-time video data originating from at least one video camera that monitors a transaction area;
receiving real-time transaction data from a transaction terminal, the real-time transaction data tracking events associated with a transaction occurring in the transaction area;
analyzing images in real-time video data to track, in substantially real-time with respect to the at least one video camera capturing the images of the transaction area, items present in the transaction area that are involved in the transaction;
classifying regions of the images in the real-time video data as representing the tracked items present in the transaction area; and
comparing the video analysis of the tracked items in the classified regions of the real-time video data to the real-time transaction data to identify a particular event that is captured by the received real-time video data but that does not have a corresponding log in the real-time transaction data indicating that the event is part of the transaction.
2. The method as in claim 1 , wherein analyzing the real-time video data further comprises:
analyzing images in a first region of interest in the real-time video data;
analyzing images in a second region of interest in the real-time video data; and
detecting occurrence of the particular event based on a combined analysis of the analysis of images captured in the first region of interest and the analysis of images captured in the second region of interest.
3. The method as in claim 2 , wherein the particular event indicates presence of a particular item in the transaction area.
4. The method as in claim 3 further comprising:
in response to detecting that the real-time video data captures presence of the particular item in the transaction area but that the real-time transaction data does not include the corresponding log indicating that the particular item in the transaction area is part of the transaction, generating a notification of the particular event.
5. The method as in claim 1 further comprising:
in response to detecting occurrence of the particular event, transmitting a notification of the particular event to security personnel.
6. The method as in claim 1 , wherein analyzing the images and classifying the regions of the real-time video data further comprises:
maintaining a reference image of the transaction area, the reference image representing background objects in the transaction area; and
comparing the images in the real-time video data to the reference image to detect an object of interest in the transaction area.
7. The method as in claim 6 , wherein maintaining the reference image further comprises:
calculating motion of the items in the transaction area; and
adapting the reference image to account for the motion of the items.
8. The method as in claim 7 , wherein comparing the images in the real-time video data to the reference image to detect the object of interest in the transaction area further comprises:
identifying the object of interest based on a difference between the images in the real-time video data and the reference image.
9. The method as in claim 6 , wherein maintaining the reference image further comprises:
estimating belt motion that occurs between frames of the video data;
calculating a translation parameter based on the belt motion; and
utilizing the translation parameter to adjust the reference image over time.