Method and system for queue length analysis
A system and method for analysing queues in frames of video enables operators to preferably draw three regions of interest overlaid upon the video as short, medium, and long queue regions that form a notional queue area within the video. The regions are drawn with knowledge of, or in anticipation of, foreground objects such as individuals and vehicles waiting for service in a queue. Examples include retail point of sale locations or for automated teller machine (ATM) transactions. In conjunction with a video analytics system that analyses the movement of the foreground objects relative to the queue regions, the system determines the number of objects occupying each queue region, length of the queue, and other queue-related statistics. The system can then create reports and send messages that include the queue analysis results for directing operators to change their staffing resources as part of a real-time queue servicing and optimization response.
1. A method of operation of a finite state machine of a video analytics system for determining whether an object has entered or exited regions in frames of video data, the method comprising: defining the regions of the frames, wherein the regions are queue regions forming a queue area; determining that the object has transitioned to a state indicating that the object has exited the regions, when the object has a tracking mask in the next frame of video data that does not overlap with any of the regions; determining that the object has transitioned to a state indicating that the object has entered an identified region of the regions, when the object has a tracking mask in the next frame of video data that overlaps with the identified region; and determining a length of a queue by successively determining if each of the regions is occupied by calculating a union, for each of the regions, of marked areas of intersection of the objects upon the queue regions; and comparing the union of the marked areas of intersection of the objects belonging to each of the regions, to a minimum occupancy area for each of the regions.
2. The method of claim 1 , further comprising initially assigning the object to an unknown state, and transitioning the object to an out-of-region state when the tracking mask of the object does not overlap with any of the regions.
3. The method of claim 2 , further comprising determining that the object remains in the unknown state when the object does not have a bounding box in a next frame of the video data.
4. The method of claim 1 , wherein overlap of a tracking mask with a region is determined when the tracking mask intersects with the region by at least a predetermined amount.
5. The method of claim 1 , further comprising determining a number of objects that are within an area by counting the objects that belong to the one or more queue regions forming the queue area.
6. The method of claim 1 , further comprising enabling drawing of the regions over video data.
7. The method of claim 1 , wherein the regions are rectangular.
8. The method of claim 1 , wherein the regions are trapezoidal.
9. A video analysis system for monitoring queues, comprising: at least one video camera generating video data of a monitored area; and a video analytics system that executes a finite state machine for determining whether an object has entered or exited regions in frames of video data, by defining the regions of the frames wherein the regions are queue regions forming a queue area, determining that the object has transitioned to a state indicating that the object has exited the regions, when the object has a tracking mask in the next frame of video data that does not overlap with any of the regions, determining that the object has transitioned to a state indicating that the object has entered an identified region of the regions, when the object has a tracking mask in the next frame of video data that overlaps with the identified region; and determining a length of a queue by successively determining if each of the regions is occupied by calculating a union, for each of the regions, of marked areas of intersection of the objects upon the queue regions and comparing the union of the marked areas of intersection of the objects belonging to each of the regions, to a minimum occupancy area for each of the regions.
10. The system of claim 9 , wherein overlap of a tracking mask with a region is determined when the tracking mask intersects with the region by at least a predetermined amount.
11. The system of claim 9 , wherein the video analytics system determines a number of objects that are within an area by counting the objects that belong to the one or more queue regions forming the queue area.
12. The system of claim 9 , wherein the video analytics system enables an operator to draw the regions over video data.
13. A method of operation of a finite state machine of a video analytics system for determining whether an object has entered or exited regions in frames of video data, the method comprising: for each frame of the frames of video data, identifying a tracking mask associated with the object in a current frame and determining whether the object remains in a current state or transitions to a new state based on the tracking mask for the object in a next frame, including: determining that the object remains in or has transitioned to an unknown state when the object does not have a bounding box in the next frame; determining that the object remains in or has transitioned to a state indicating that the object has exited the regions, when the object has a tracking mask in the next frame that does not overlap with any of the regions by at least a predetermined amount; determining that the object remains in or has transitioned to a state indicating that the object has entered an identified region of the regions when the object has a tracking mask in the next frame that overlaps with the identified region by at least the predetermined amount, determining whether a particular region is occupied by marking, for each object that was determined to have entered the particular region, an area of intersection between the particular region and the tracking mask for the object, and determining a length of a queue by successively determining if each of the regions is occupied by calculating a union of all marked areas of intersection for the particular region, and determining that the region is occupied in response to determining that the union of the marked areas of intersection for the particular region is not less than a predetermined minimum occupancy area for the particular region.
14. The method of claim 13 , further comprising initially assigning the object as being in the unknown state.
15. The method of claim 13 , wherein the predetermined amount is defined by an operator of the video analytics system.