| Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion -> Monitor Keywords |
|
|
 
Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motionRelated Patent Categories: Image Analysis, Pattern RecognitionMethod and apparatus for maintaining a background image model in a background subtraction system using accumulated motion description/claimsThe Patent Description & Claims data below is from USPTO Patent Application 20070177800, Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion. Brief Patent Description - Full Patent Description - Patent Application Claims
FIELD OF THE INVENTION [0001] The present invention relates generally to imaging processing techniques, and, more particularly, to techniques for generating and maintaining background model images in a background subtraction system. BACKGROUND OF THE INVENTION [0002] Background subtraction is a common technique for detecting moving objects in a largely stationary environment. Generally, background subtraction techniques compare a current image against a reference "empty" image and note regions of change. Such differences generally correspond to the moving foreground objects of interest for applications such as surveillance. However, a background model must first be created and then maintained. [0003] A background model can be created, for example, by memorizing an "empty" image where there are no visible objects. However, this method is not generally applicable because it is difficult to ensure that a scene is clear of all moving objects, especially if it is being remotely monitored. Another technique monitors an incoming video stream for motion energy by subtracting adjacent frames. When the overall energy is low enough, an input frame is captured and thereafter used as the background model. Unfortunately, if a moving object, such as a pedestrian, momentarily pauses, a background image might be acquired which erroneously includes this temporarily immobile object. [0004] Conversely, moving objects may enter a scene and then stop moving (e.g., a parked car). Similarly, if a person enters a room and then becomes immobile, such as taking a nap, the person persists as a detected object. In many cases, such objects would be better interpreted as a part of the "new" background. Moreover, if a person is present when the reference "empty" image is acquired, the person will be detected once he or she starts to move across the scene. However, even after the person has completely exited the scene, a "hole" where the person was originally positioned will likely continue to be perceived as an object. [0005] The background updates can be significant events in themselves. If an item is introduced into the scene (such as a briefcase carried surreptitiously by some agent), the item will also be marked as part of the foreground despite having no motion itself. A related situation is where some object that was part of the original reference image (e.g., a laptop computer) is removed from the scene. There will be a difference "hole" left behind in this case that is not only non-moving, but also not a solid object. There are several methods that can be used to locate non-moving regions in a background subtraction system, but it is difficult to classify such regions as abandoned objects, removed objects, state changes, or another phenomena. [0006] The region type can be determined, for example, by comparing the pixel pattern in the region to the template of some known object (e.g., a computer monitor). If there is a non-moving foreground region and the associated pixels match the template, then a deposit event has occurred, otherwise a removal event is recorded. However, this approach requires that the system have a number of templates for each kind of object it cares about. Typically, such models must be manually entered, or an operator has to at least mark the boundaries of some region. A related approach requires that the system know something about the background near the non-moving region. If, for instance, the room had uniformly green walls, then the interior of the non-moving region could be examined to see if it was green or not (implying a removal or a deposit, respectively). Again, this is not a general purpose solution and works best if the characteristics of the environment can be chosen at will (like applying green paint). [0007] Updating the background model is particularly important for environments in which lighting changes over time (e.g., outdoors). Otherwise, lighting changes, such as the sun coming out from behind clouds, can cause large areas of the image to be falsely declared as foreground as they become better illuminated. Updating the background model is often done by slowly blending in newly acquired images with the old model. However, if the blending rate is fast and applied to the whole image, moving objects start to leave ghostly trails across the background model. Such anomalies can cause the system to both falsely detect nonexistent objects and miss detecting some valid objects. If, on the other hand, the blending rate is set very slow to reduce this effect, the system might not adapt quickly enough to the types of lighting changes that are present. Another updating option is to periodically reinitialize the background model from scratch using the current video frame. This might be done, for instance, either on a regular schedule or when too large a portion of the image is marked as foreground. While this approach prevents ghosting, it still suffers from the same initial model acquisition problems as described above. [0008] Background blending, as discussed above, will eventually "erase" all stationary objects as well as "holes" left by removed objects. Unfortunately, this method tends to erase objects by shrinking them, leaving invalid partial objects as the regions are absorbed into the background. Moreover, background blending can also leave "ghost" objects behind if, say, a moving person lingers too long in one area. There exist more sophisticated systems that model the intensity at each pixel as one of several Gaussian distributions. These systems have better immunity to the "ghosting" problem, but the decision to switch from one Gaussian model to another is typically made independently for each pixel. This leads to objects or holes "sparkling" out, with incorrect ragged objects detected during the transition. [0009] As apparent from the above-described deficiencies with conventional techniques for generating and updating a background model, a need exists for methods and apparatus for improved techniques for generating and updating a background model. A further need exists for methods and apparatus for visual background subtraction that address each of the above-identified problems using one or more software preprocessing modules. SUMMARY OF THE INVENTION [0010] Generally, methods and apparatus are provided for maintaining a background image model in a background subtraction system using accumulated motion. According to one aspect of the invention, a background image model is maintained by obtaining a map of accumulated motion; and adjusting the background image model based on the map of accumulated motion. In one implementation, a counter is maintained indicating a degree of motion in a portion of a sequence of images. [0011] The map of accumulated motion may be obtained, for example, based on one or more of motion field images; stability maps; frame differences; or information from a background subtraction system. Based on the map of accumulated motion, objects can be added to or removed from the background model or the background model can be otherwise updated. [0012] According to another aspect of the invention, one or more pixels from an image are added to the background image model if a stability measure for the one or more pixels satisfies a predefined criteria. In addition, a portion of the background image model can be invalidated in regions where the map of accumulated motion exceeds a predefined threshold. A portion of an image can be copied, for example, to the background image model if a stability measure satisfies a predefined criteria. [0013] A more complete understanding of the present invention, as well as further features and advantages of the present invention, will be obtained by reference to the following detailed description and drawings. BRIEF DESCRIPTION OF THE DRAWINGS [0014] FIG. 1 is a schematic block diagram of a background model processing system incorporating features of the present invention; [0015] FIG. 2 is a flow chart describing an exemplary implementation of a background building method that may be employed by the background building module of FIG. 1; [0016] FIG. 3 is a flow chart describing an exemplary implementation of a background invalidation method that may be employed by the background invalidation module of FIG. 1; [0017] FIG. 4 is a flow chart describing an exemplary implementation of a background region removal process that may be employed by the background region removal module of FIG. 1; and [0018] FIG. 5 is a flow chart describing an exemplary implementation of a foreground blending module that may be employed by the foreground blending module of FIG. 1. DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS [0019] The present invention provides methods and apparatus for maintaining a background image model in a background subtraction system using a map of accumulated motion. As discussed hereinafter, the map of accumulated motion can be obtained, for example, based on frame differences or from a background subtraction system. The accumulated motion can be employed to add or remove objects from the background model or to update the background model. The background model can be updated on a pixel-by-pixel basis or at a larger object level. Continue reading about Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion... Full patent description for Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion Brief Patent Description - Full Patent Description - Patent Application Claims Click on the above for other options relating to this Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion patent application. ###  How KEYWORD MONITOR works... a FREE service from FreshPatents1. Sign up (takes 30 seconds). 2. Fill in the keywords to be monitored. 3. Each week you receive an email with patent applications related to your keywords. Start now! - Receive info on patent apps like Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion or other areas of interest. ### Previous Patent Application: Image analysis Next Patent Application: Constraint-based ordering for temporal coherence of stroke-based animation Industry Class: Image analysis ### FreshPatents.com Support Thank you for viewing the Method and apparatus for maintaining a background image model in a background subtraction system using accumulated motion patent info. IP-related news and info Results in 0.26978 seconds Other interesting Feshpatents.com categories: Computers: Graphics , I/O , Processors , Dyn. Storage , Static Storage , Printers 174 |
 * Protect your Inventions * US Patent Office filing 
PATENT INFO |
|
