Anomaly detection in images and videos

• US 8,724,904 B2
• Filed: 10/25/2011
• Issued: 05/13/2014
• Est. Priority Date: 10/25/2011
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Fujiki, Yuichi, Haas, Norman, Li, Ying, Otto, Charles A., Paluri, Balamanohar, Pankanti, Sharathchandra
• Primary Examiner(s): Koziol, Stephen R
• Assistant Examiner(s): Azima, Shaghayegh
• Application Number:
  US13/280,896
• Publication Number:
  US 20130101221A1
• Time in Patent Office: 931 Days
• Field of Search: 382/195, 382/225, 382/104
• US Class Current: 382/195
• CPC Class Codes: B61L 23/044 Broken rails
• CPC Class Codes: B61L 23/047 Track or rail movements
• CPC Class Codes: B61L 23/048 Road bed changes, e.g. road bed erosion
• CPC Class Codes: G06F 18/23 Clustering techniques
• CPC Class Codes: G06F 18/2433 Single-class perspective, e.g. one-against-all classification; Novelty detection; Outlier detection
• CPC Class Codes: G06V 10/762 using clustering, e.g. of similar faces in social networks

Abstract

A system, method, and computer program product for detecting anomalies in an image. In an example embodiment the method includes partitioning each image of a set of images into a plurality of image local units. The method further includes clustering all local units in the image set into clusters, and consequently assigning a class label to each local unit based on the clustering results. The local units with identical class labels having at least one substantially related image feature. Further, the method includes assigning a weight to each of the local units based on a variation of the class labels across all images in a set of images. The method further includes performing a clustering over all images in the set by using a distance metric that takes the learned weight of each local unit into account, then determining the images that belong to minorities of the clusters as anomalies.

Claims

1. A non-transitory computer program product for detecting tie plate anomalies in a tie plate image, the computer program product comprising:

a computer readable storage medium having computer readable program code embodied therewith, the computer readable program code configured to;

localize a tie plate region in an image from a set of tie plate images;

divide the tie plate region into a set of local units;

extract features from each local unit in the set of local units;

perform a clustering over all local units;

assign a class label to each local unit, the class label indicates a semantic content of the local unit;

determine the weight of each local unit across all images in the set of images; and

categorize the tie plate image as anomalous based on the weight assigned to the local units.

2. The computer program product of claim 1, wherein the computer readable program code is further configured to represent each tie plate region from the set of images with concatenated unit labels, wherein the concatenated unit labels are feature vectors representing the tie plate region.

3. The computer program product of claim 1, wherein the computer readable program code is further configured to perform a global clustering over all tie plate regions in the set of images based on a distance calculation between images in the image set, wherein the distance calculation is based, in part, on the weight of each of the local units in each tie plate region.

4. The computer program product of claim 3, wherein the computer readable program code is further configured to determine the tie plate regions that belong to classes of minorities from the global clustering are anomalous.

9. The computer program product of claim 1, wherein the local units form an M by N two-dimensional regular grid having M rows and N columns.

10. The computer program product of claim 1, wherein the computer readable program code is further configured to:

calculating a centroid for each cluster of the local units; and

incrementally update the centroid of each local unit cluster in real time by including the corresponding local units of the new tie plate images.

11. The computer program product of claim 10, wherein the computer readable program code is further configured to incrementally update a weight of each of the local units in real-time by utilizing an original weight and an alternate weight array.

5. A system for detecting tie plate anomalies in tie plate images, the system comprising:

a processor;

a memory coupled to the processor, the memory having computer readable program code embodied therewith, the computer readable program code configured to;

localize a tie plate region in an image from a set of tie plate images;

divide the tie plate region into a set of local units;

extract features from each local unit in the set of local units;

perform a clustering over all local units;

assign a class label to each local unit, the class label indicates a semantic content of the local unit;

determine the weight of each local unit across all images in the set of images; and

categorize the tie plate image as anomalous based on the weight assigned to the local units.

6. The system of claim 5, wherein the local units form an M by N two-dimensional regular grid having M rows and N columns.

7. The system of claim 5, wherein the computer readable program code are further configured to:

calculate a centroid for each cluster of the local units; and

incrementally update the centroid of each local unit cluster in real time by including the corresponding local units of the new tie plate images.

8. The system of claim 7, wherein the computer readable program code are further configured to incrementally update a weight of each of the local units in real-time by utilizing an original weight and an alternate weight array.