Methods and Apparatus for Item Identification Using Brightness Compensation

• US 20150213326A1
• Filed: 01/28/2014
• Published: 07/30/2015
• Est. Priority Date: 01/28/2014
• Status: Active Grant

• Current Assignee: NCR Voyix Corporation
• Original Assignee: NCR Corporation
• Inventors: Utsumi, Yoshitaka
• Granted Patent: US 9,299,007 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06V 10/28 Quantising the image, e.g. histogram thresholding for discrimination between background and foreground patterns
• CPC Class Codes: G06V 10/56 relating to colour
• CPC Class Codes: H04N 23/80 Camera processing pipelines; Components thereof

Abstract

Methods and apparatus are provided for low resolution item identification using brightness compensated images. An image of an unknown item is captured in digital RGB image format and converted to HSV image format. Then the HSV formatted image is quantized which greatly reduces the resolution of the image. The quantized HSV image data of the unknown item is compared to a plurality of quantized HSV image data for known items. The comparison includes using a signal-to-noise ratio calculated using the quantized HSV image data for both the unknown and known items. A match is found when the calculated signal-to-noise ratio is above a predetermined threshold value.

Claims

1. A computer implemented method to identify an unknown item presented to an optical code scanner based on the appearance of the item, the method comprising:

capturing an RGB formatted digital image of the unknown item by the optical code scanner;

converting the captured image from RGB format to HSV format;

generating a transform from the HSV formatted image using the hue and saturation data and a weighted amount of the brightness value in the HSV formatted image; and

comparing the transform to data for a plurality of transforms for known items to determine the identity of the unknown item.

2. The method of claim 1 wherein the transform is generated by quantizing the converted HSV image to reduce the amount of data that represents the image of the unknown item and wherein comparing the transform includes comparing the quantized HSV image data for the unknown item to a plurality of quantized HSV image data for known items where the comparing includes using a calculated signal to noise ratio (SNR) between the unknown and known quantized image data to determine when a match is found between the unknown item and a known item.

3. A computer implemented method to identify an unknown item presented to an optical code scanner based on the appearance of the item, the method comprising:

capturing an RGB formatted digital image of the unknown item by the optical code scanner;

converting the captured image from RGB format to HSV format;

quantizing the converted HSV formatted image using the hue and saturation data and a weighted amount of the brightness value in the HSV formatted image wherein the weighted brightness value reduces variations in the brightness of the unknown item; and

comparing the quantized image data for the unknown item to a plurality of similarly quantized image data for known items to determine a match between the quantized unknown item and a quantized known item.

4. The method of claim 3 wherein determining a match includes using a calculated signal to noise ratio (SNR) between the unknown and known quantized image data to determine when a match is found between the unknown item and a known item.

5. The method of claim 3, wherein the quantizing step includes performing a pixelation on the captured image data to reduce the size of original image data.

6. The method of claim 3, wherein the comparing step includes calculating the root-mean-square (RMS) for the quantized image data of the unknown item.

7. The method of claim 6, wherein the comparing step further includes generating variance image data where the variance image data is the difference between the quantized image data of the unknown item and quantized image data for the known item.

8. The method of claim 7, wherein calculating the SNR includes calculating a signal value which is the RMS value of the quantized image data of the known item and calculating the “

noise”

value which is the RMS value of the variance image data and the value for the SNR is calculated by dividing the “

signal”

value by the “

noise”

value.

9. The method of claim 8, wherein a predetermined threshold SNR value is used to determine the minimum confident level that can be used to determine that a match has been found.

10. The method of claim 8, wherein within the quantized image data for both the known and unknown items, the data for each pixel is reduced to include data for only one color.

11. The method of claim 3, wherein the plurality of quantized image data for known items are stored in a database.

12. The method of claim 11, wherein the database is stored in the optical code scanner.

13. The method of claim 11, wherein the database is a customized database and accepts the quantized image data for the unknown item and performs the comparison to the quantized image data for the known items stored in the database and returns the known item that matches the unknown item.

14. The method of claim 3, further including selecting the plurality of quantized image data for known items from a larger set of quantized image data for known items using one or more physical attributes of the unknown item.

15. The method of claim 3, wherein the quantizing step reduces the number of pixels represented in the captured image data down to fewer than 200 pixels.

16. The method of claim 3, wherein there are multiple quantized images for each known item where at least one quantized image was generated from an image taken of the front of the item and at least one quantized image was generated from an image taken of the back of the item.

17. An optical code scanner comprising:

an image capture device;

a memory including computer instructions; and

a processor in communication with the image capture device and the memory where the computer instructions when executed by the processor cause the processor to perform functions including;

capturing an RGB formatted digital image of the unknown item by the image capture device;

converting the captured image from RGB format to HSV format;

quantizing the converted HSV formatted image using the hue and saturation data and a weighted amount of the brightness value in the HSV formatted image wherein the weighted brightness value reduces variations in the brightness of the unknown item; and

comparing the quantized image data for the unknown item to a plurality of similarly quantized image data for known items where the comparing includes using a calculated signal to noise ratio (SNR) between the unknown and known quantized image data to determine when a match is found between the unknown item and a known item.

18. The optical code scanner of claim 17, wherein the quantizing step includes performing a pixelation on the captured image data to reduce the size of original image data.

19. The optical code scanner of claim 18, wherein the comparing step includes calculating the root-mean-square (RMS) for the quantized image data of the unknown item.

20. The optical code scanner of claim 18, wherein the comparing step further includes generating variance image data where the variance image data is the difference between the quantized image data of the unknown item and quantized image data for the known item.

21. The optical code scanner of claim 20, wherein calculating the SNR includes calculating a signal value which is the RMS value of the quantized image data of the known item and calculating the noise value which is the RMS value of the variance image data and the value for the SNR is calculated by dividing the signal value by the noise value.

22. The optical code scanner of claim 21, wherein a predetermined threshold SNR value is used to determine the minimum confident level that can be used to determine a match has been found.

23. The optical code scanner of claim 17, wherein the plurality of quantized image data for the known items are stored in a database stored in the memory.