| US 7,526,140 B2 | ||
| Model-based localization and measurement of miniature surface mount components | ||
| Ming Fang, Princeton Jct., N.J. (US); and Jenn-Kwei Tyan, Princeton, N.J. (US) | ||
| Assigned to Siemens Corporate Research, Inc., Princeton, N.J. (US) | ||
| Filed on Mar. 04, 2005, as Appl. No. 11/72,220. | ||
| Application 11/072220 is a division of application No. 10/042887, filed on Jan. 09, 2002, granted, now 6,980,685, filed on Dec. 27, 2005. | ||
| Claims priority of provisional application 60/263293, filed on Jan. 22, 2001. | ||
| Prior Publication US 2005/0169512 A1, Aug. 04, 2005 | ||
| Int. Cl. G06K 9/40 (2006.01); G06K 9/32 (2006.01) | ||
| U.S. Cl. 382—266 [382/300; 348/441; 358/525; 708/290] | 11 Claims |
| 1. A computer readable medium embodying a program of instructions executable by machine to perform steps for automatically
detecting nodules from image data, the steps comprising:
receiving an image having an object corresponding to an object type;
iteratively segmenting the object;
applying a moment transformation to the segmented object;
measuring the object, wherein measuring the object comprises: detecting and interpolating edges of the object; and iteratively
optimizing results of the measurement;
offline modeling of the object type; and
runtime matching of the object corresponding to the object type, wherein runtime matching comprises: receiving the image having
the object corresponding to the object type; performing a coarse search for the object; and performing a refined search for
the object,
wherein performing a coarse search comprises localizing the object from the image in accordance with a model, and wherein
performing a refined search comprises measuring the localized object,
wherein localizing comprises: iteratively segmenting the object; and applying the moment transformation to the segmented object,
and
wherein iteratively segmenting the object comprises:
selecting an initial estimate of a threshold by using the average gray-level of the 2n brightest pixels in the image, where
n is the size of the model;
segmenting the image into a background region and an object region in accordance with the threshold, with the pixels having
a gray-level less than the threshold being assigned to the background region and all other pixels being assigned to the object
region;
calculating the mean gray-levels within the background and object regions, respectively;
calculating a new threshold in accordance with the calculated mean gray-levels and the number of pixels in each region;
repeating the above steps of segmenting, calculating gray-levels and calculating new thresholds until convergence is reached;
and
obtaining the segmented object from the final pixels in the object region.
|