Method to aid object detection in images by incorporating...

Details Method to aid object detection in images by incorporating... Method to aid object detection in images by incorporating...

2000-02-25

2003-06-10

Davis, George B. (Department: 2121)

Data processing: artificial intelligence

Fuzzy logic hardware

Analog fuzzy computer

C706S004000, C706S008000, C706S009000

Reexamination Certificate

active

06578017

ABSTRACT:

BACKGROUND OF THE INVENTION
Automated object detector (AOD) systems use computer-based algorithms to detect objects of interest in an image. AODs typically measure how well each piece of an image represents a known object, then highlight those pieces where the measure exceeds some threshold. In theory, an AOD can scan a complex image to identify objects better and faster than a human can. In practice however, the simple comparisons performed by AODs are not very robust, and to avoid missing valid objects, they often detect questionable objects, resulting in false detections. A better approach, as described below, would be to encode, either manually via direct human input or automatically by a computer system, available contextual information concerning an image in order to filter out false alarms while increasing the probability that the AOD correctly detects valid objects. There is also the need to combine contextual information with an image to aid the human in detecting objects of interest when no AOD is available.
BRIEF SUMMARY OF THE INVENTION
The present invention uses fuzzy logic and/or probability distributions to automatically calculate and display the effects of contextual information (location and characteristics of “context objects”) on confidence that an object in an image is an object of interest (“target object”). The goal is to assist in determining the location and type of target objects of interest in that imagery—both for AOD operation and for human-only (no AOD) operation. The invention's uses include post-processing imagery to enhance the validity of the results of an AOD system or human detection, and pre-processing imagery to enhance the effectiveness of an AOD system or human detection. The imagery used by the invention can come from any kind of imaging sensor (local or remote to the AOD system or human), or can be non-sensor imagery (e.g., two-and three-dimensional maps), and can be live (real-time) or archived imagery. The locations of context objects can be provided by a human or a computer. The resulting set of data, including the original imagery, the locations of context objects, any results from an AOD, and predictions about target object type and location, can be combined by the current invention into a display that helps a human better understand where target object appear in the imagery. This invention can help resolve conflicting versus corroborating contextual information and evidence, positive versus negative types of evidence, vagueness and uncertainty, spatially and/or temporally changing information, multiple types of evidence—including information separated in space and time, and attention competing (distracting) evidence. The context objects can be either visible to a human or not visible to a human. They can be used to provide a reference point to highlight regions of interest in the image, a placeholder to return to after searching an image, and/or a reference of known size as a yardstick to determine size and distances of context objects and targets in an image.
Has many APPLICATIONS. Other modifications are also possible. For example, as discussed previously, the invention advantageously may be employed in other applications such as, for example, medical imaging. The invention also advantageously may be used for remote sensing and prospecting, geological fault detection, weather forecasting, etc., aiding decision making in dynamic social and economic systems, and in general any application where a human must make decisions in a time-critical fashion.


REFERENCES:
patent: 5341439 (1994-08-01), Hsu
patent: 5436979 (1995-07-01), Gray et al.
patent: 5648782 (1997-07-01), Albo et al.
patent: 6072889 (2000-06-01), Deaett et al.
patent: 5-290169 (1993-11-01), None
patent: 5-290170 (1993-11-01), None

Affiliated with

Also associated with

Rating

Method to aid object detection in images by incorporating... does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Method to aid object detection in images by incorporating..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method to aid object detection in images by incorporating... will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-3161830