Image analysis – Applications – Vehicle or traffic control
Reexamination Certificate
1999-07-14
2003-04-29
Mehta, Bhavesh (Department: 2721)
Image analysis
Applications
Vehicle or traffic control
C382S209000, C382S218000, C382S227000, C382S278000, C382S302000
Reexamination Certificate
active
06556692
ABSTRACT:
CROSS-REFERENCE TO RELATED APPLICATIONS
Priority is claimed with respect to German application No. 198 31 413.2 filed in Germany on Jul. 14, 1998, the disclosure of which is incorporated herein by reference.
BACKGROUND OF THE INVENTION
The invention relates to an image-processing method, and an apparatus for carrying out the method, in which one or a plurality of feature image patterns is correlated with an original image, with a feature image scene being formed from an original image scene through feature extraction, the feature values of the scene being locally softened through a transformation or a filter, and the transformed image scene being correlated with the feature image patterns, and/or the feature values of the feature image patterns being softened through a transformation or a filter, and the transformed feature image patterns being correlated with the feature image scene, and with objects being detected and recognized with the processed image data.
The use of the invention is not limited to the recognition of objects in traffic, but can be used in any type of image-monitoring systems and in image databases.
A publication by W. J. Rucklige in the Int. J. of Computer Vision 24(3), pp. 251-270 (1997) discloses a method of correlating a predetermined pattern with an image to be evaluated using a distance transformation; in this method, a pattern is first subjected to an affine transformation with coefficients to be determined, and correlated afterward with the image.
In the IEEE Trans. on Pattern Analysis and Machine Intelligence, 10(6), pp. 849-865 (1988), G. Borgefors describes an image-recognition method based on distance transformations, which method is expanded by the use of an image-resolution pyramid.
In “Hybrid approach for traffic sign recognition” by Janssen et al. in the Proc. of Intelligent Vehicle Conference, pp. 390-395 (1993), a method is described that employs adjacent color fields to detect regions in the image that may include traffic signs. The quantity of regions to be analyzed is further limited by the evaluation of heuristics based on format attributes of the regions (e.g. surface, eccentricity) and proximity conditions. The image regions that remain as possible traffic signs are cut out of the original image and classified.
A method of recognizing traffic signs is disclosed in G. Piccioli et al., “Image and Computing,” No. 14, pp. 209-223 (1996). In a color-based setup for recognizing traffic signs, different recognition possibilities are disclosed, with which possible traffic sign shapes (circles, triangles) are evaluated. First, the contour segments are detected; then, it is checked whether these can be combined to yield the desired shape.
“Fast Object Recognition in Noisy Images Using Simulated Annealing” by M. Betke et al. in Proc. of IEEE Int. Conf. on Computer Vision, pp. 523-530 (1995) describes a pictogram-based method of recognizing traffic signs, in which different patterns are correlated with an image. The patterns contain the pictogram information of searched traffic signs.
The patent publication DE 36 19 824 C2 by E. D. Dickmanns describes an apparatus for indicating the maximum speed for road vehicles that is presently dictated or safe, depending on the environmental conditions. A pivotable camera whose images are evaluated is employed to detect and recognize traffic signs with the aid of an information and evaluation device. The result of the recognition is combined with other sensor information in a matching circuit, then displayed for the driver.
SUMMARY OF THE INVENTION
It is an object of the invention to optimize the capability of object detection in images based on known methods.
The above and other objects are accomplished in the context of the image processing method first mentioned above, wherein during the correlation of the transformed image scenes with the feature image patterns, knowledge is applied that is based on the knowledge about the objects to be detected, and/or on global knowledge about the situation in the scenario associated with the original image, and, based on this knowledge, only specific regions of the feature image scene and/or only specific feature image patterns are correlated with one another.
An advantage of the invention is that, through the correlation of the feature images of an image scene with feature images of patterns (or vice versa), the feature images having been locally softened through transformation, it becomes possible to detect and/or recognize objects in natural scenes with the use of images. The objects can be partially covered and appear in numerous forms in the image (e.g. appear different due to various light conditions and different shapes, textures, positions, orientations and movements of the object). This is the case, for example, for objects occurring in traffic, such as pedestrians, vehicles, traffic signs and the like.
The feature image patterns of the objects to be detected and recognized are determined analytically through the sampling of examples, or a priori knowledge about the objects.
In the invention, the special correlation performed in the image cutout, which involves the feature images of an image scene that have been locally softened due to the transformation and the feature images of patterns (or vice versa), effects a successful detection and recognition of objects, even if data are absent or do not correspond to a so-called prototype. The method of locally softening the features, which is used in the procedure prior to the correlation, permits necessary tolerances between the pattern and the form in which the object appears.
Another advantage of the method of the invention is that N pattern images representing one or more objects to be detected are correlated in an image scene through the creation of a pattern tree structure (pattern hierarchy). The method is thus more efficient than methods in which N respective patterns are correlated separately.
Independently of the use of a pattern tree structure, the method of the invention has the advantage that a plurality of feature types, such as edge orientations, can be taken into consideration simultaneously in the correlation. The use of M different feature types reduces the probability of an erroneous pattern allocation in the image. The correlation is performed with corresponding features.
A further advantage of the invention is that the detection results are evaluated with known classification methods of pattern recognition, and an object recognition is thus performed.
Because the method is not necessarily based on the evaluation of color information, an apparatus having a monochromatic camera can advantageously be used; in this instance, the outlay for the data conversion is low, making the apparatus inexpensive.
REFERENCES:
patent: 5123057 (1992-06-01), Verly et al.
patent: 5619596 (1997-04-01), Iwaki et al.
patent: 6181805 (2001-01-01), Koike et al.
patent: 6307964 (2001-10-01), Lin et al.
Gälich, R.-E; Ritter, W. “Aknowledge based system for traffic sign recognition,”Mustererkennung [Pattern Recognition]21990, 12thDAGM Symposium, Oberkochen, Sep. 24-26, 1990, Springer Publishing House, pp. 82-89.
K.-D. Kuhnert, A. Zapp: “Wissensgesteuerte Bildfolgeauswertung Zur Automatischen Führung Von Strassenfahrzeugen in Echtzeit” [Knowledge-Controlled Picture Sequence Evaluation for the Automatic Guidance in Real Time of Vehicles on the Road],Mustererkennung[Pattern Recognition ]1985, 7thDAGM Symposium, Erlangen, Springer Publishing House, pp. 102-106.
U. Büker, H. Austermeier, G. Hartmann, B. Mertsching “Verkehrszeichenanalyse in Hierarchisch Codierten Bildern” [Traffic Sign Analysis in Hierarchically Coded Images],Mustererkennung[Pattern Recognition ]1993, 15thDAGM Symposium, Lübeck, Springer Publishing House, pp. 694-701.
D.P. Huttenlocher, G. A. Klanderman, W. J. Rucklidge “Comparing Images Using the Hausdorff Distance” [IEEE Transactions on Pattern Analysis and Machine Intelligence ]vol. 15, No. 9, Sep. 1993, pp. 850-862.
P. Seutens, P. Fau, A. J. Ha
Daimler-Chrysler AG
Davidson Davidson & Kappel LLC
Desire Gregory
Mehta Bhavesh
LandOfFree
Image-processing method and apparatus for recognizing... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Image-processing method and apparatus for recognizing..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Image-processing method and apparatus for recognizing... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3070315