Image selection and recognition processing from a video feed

Patent metadata

Field	Value
Publication number	US-9626577-B1
Application number	US-201414486509-A
Country	US
Kind code	B1
Filing date	Sep 15, 2014
Priority date	Sep 15, 2014
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

A system that selects image frames from a video feed for recognition of objects (such as physical objects, text characters, or the like) within the image frames. The individual frames are selected using robust historical metrics that compare individual metrics of the particular image (such as focus, motion, intensity, etc.) to similar metrics of previous image frames in the video feed. The system will select the image frame for object recognition if the image frame is relatively high quality, that is the image frame is suitable for a later object recognition processing.

First claim

What is claimed is:

1. A computing system configured to select an image for object recognition, the system comprising:
at least one processor;
memory including instructions which, when executed by the at least one processor, cause the system to perform a set of actions comprising:
receiving a plurality of sequential images in a camera video feed, the plurality of sequential images comprising a first image and preceding images, the preceding images occurring in the sequential images prior to the first image;
determining a respective value for a focus metric for each image of the plurality of sequential images, wherein the focus metric measures a respective focus level of each image;
determining a first focus metric value for the first image;
calculating a historical focus metric value based on a plurality of respective focus metric values for the preceding images;
determining a respective value for a motion metric for each image of the plurality of sequential images, wherein the motion metric measures a respective motion level of each image;
determining a first motion metric value for the first image;
calculating a historical motion metric value based on a plurality of respective motion metric values for the preceding images;
computing a suitability metric value by processing the first focus metric value, the historical focus metric value, the first motion metric value, and the historical motion metric value using a classifier model, wherein the suitability metric value indicates a suitability for performing optical character recognition on the first image; and
selecting the first image for optical character recognition based at least in part on the suitability metric value.

2. The computing system of claim 1 , the set of actions further comprising:
identifying a glyph region within the first image, wherein the glyph region includes text characters for optical character recognition;
determining a region focus metric value, wherein the region focus metric value measures a focus level in the glyph region;
determining a region motion metric value, wherein the region motion metric value measures an amount of motion detected in the glyph region;
processing the region focus metric value and region motion metric value using a second classifier model to obtain a second suitability metric value, wherein the second suitability metric value indicates a suitability for performing optical character recognition on the glyph region;
comparing the second suitability metric value to a second threshold,
wherein the selecting is further in response to the second suitability metric value being above the second threshold.

3. A computer-implemented method comprising:
receiving sequential images in a camera image feed, the sequential images comprising a first image and preceding images, the preceding images occurring in the sequential images prior to the first image;
determining a first focus value for the first image;
determining a plurality of respective focus values for at least a subset of the preceding images;
determining a historical focus value based on the plurality of respective focus values for at least the subset of the preceding images;
determining a first image quality value for the first image, the first image quality value based on one or more of: a first motion value, a first intensity value, a first sharpness value, or a first contrast value;
determining a plurality of respective second image quality values for at least the subset of the preceding images, a respective second image quality value based on one or more of: a respective second motion value, a respective second intensity value, a respective second sharpness value, or a respective second contrast value;
determining a historical image quality value based on the plurality of respective second image quality values for at least the subset of the preceding images;
processing at least the first focus value, the historical focus value, the first image quality value, and the historical image quality value using a classifier; and
selecting the first image for at least one of character recognition processing or object recognition processing based on an output from the classifier.

4. The computer-implemented method of claim 3 , further comprising:
identifying a glyph region within the first image, wherein the glyph region includes text characters for optical character recognition; and
determining a region quality metric value for the glyph region,
wherein the selecting is further based on the region quality metric value.

5. The computer-implemented method of claim 3 , wherein the classifier uses at least one of the following techniques: support vector machines, neural networks, logistic regression, decision trees, random forest, or adaptive boosting.

6. The computer-implemented method of claim 3 , further comprising:
determining a mean value using the plurality of respective focus values; and
determining a standard deviation value using the plurality of respective focus values,
wherein the historical focus value is determined based on the mean value and the standard deviation value.

7. The computer-implemented method of claim 6 , wherein the historical focus value is the standard deviation value divided by the mean value.

8. The computer-implemented method of claim 3 , further comprising determining a suitability focus value based on the first focus value and the historical focus value, wherein:
the suitability focus value indicates a suitability for performing recognition processing on the first image; and
selecting the first image is further based on the suitability focus value.

9. The computer-implemented method of claim 8 , further comprising:
determining a second suitability focus value for a second image in the sequential images; and
comparing the suitability focus value to the second suitability focus value,
wherein selecting the first image is further based on comparing the suitability focus value to the second suitability focus value.

10. A computing device comprising:
at least one processor;
memory including instructions which, when executed by the at least one processor, cause the device to perform a set of actions comprising:
receiving sequential images in a camera image feed, the sequential images comprising a first image and preceding images, the preceding images occurring in the sequential images prior to the first image;
determining a first focus value for the first image;
determining a plurality of respective focus values for at least a subset of the preceding images;
determining a historical focus value based on the plurality of respective focus values for at least the subset of the preceding images;
determining a first image quality value for the first image, the first image quality value based on one or more of: a first motion value, a first intensity value, a first sharpness value, or a first contrast value;
determining a plurality of respective second image quality values for at least the subset of the preceding images, a respective second image quality value based on one or more of: a respective second motion value, a respective second intensity value, a respective second sharpness value, or a respective second contrast value;
determining a historical image quality value based on the plurality of respective second image quality values for at least the subset of the preceding images;
processing at least the first focus value, the historical focus value, the first image quality value, and the historical image quality value using a classifier; and
selecting the first image for at least one of character recognition processing or object recognition processing based on an output from the clas

Assignees

• Amazon Tech Inc

Inventors

• Yu Qingfeng
• Liu Yue

Classifications

• G06V10/993 Primary

  Evaluation of the quality of the acquired pattern · CPC title

• G06K2009/4666

  Physics · mapped topic

• G06K9/46

  Physics · mapped topic

• G06K9/18 Primary

  Physics · mapped topic

• G06K9/6267

  Physics · mapped topic