Method for determining an in-focus position and a vision inspection system

• US 10,210,376 B2
• Filed: 02/19/2008
• Issued: 02/19/2019
• Est. Priority Date: 03/08/2007
• Status: Active Grant

• Current Assignee: Cellavision AB
• Original Assignee: Cellavision AB
• Inventors: Hedlund, Sven, Beijbom, Oscar, Almers, Martin
• Primary Examiner(s): Torrente, Richard T
• Application Number:
  US12/449,873
• Publication Number:
  US 20100177187A1
• Time in Patent Office: 4,018 Days
• Field of Search: 348 79, 382159, 382134
• CPC Class Codes: G03B 13/32 Means for focusing
• CPC Class Codes: G06T 2207/30024 Cell structures in vitro; Tissue sections in vitro
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 20/693 Acquisition

Abstract

In one embodiment of the present invention, a method is disclosed for determining a difference between a sample position and an in-focus position, as well as a vision inspection system. In a first step image data depicting a sample is captured. Next, a feature set is extracted from the image data. Thereafter, the feature set is classified into a position difference value, corresponding to the difference between the sample position and the in-focus position, by using a machine learning algorithm that is trained to associate image data features to a position difference value.

Claims

1. A method for determining a difference between a sample position and an in-focus position based on a single image, said method comprising:

capturing image data for the single image, the image data depicting a sample when in said sample position;

determining foreground segments and background segments in said image data;

segmenting said foreground segments into segments of object classes, at least one of said object classes selected from a group including a red blood cell (RBC) object class and a white blood cell (WBC) object class;

extracting a feature set from said image data, said feature set including a plurality of contrast features, the plurality of contrast features being generated using at least a wavelet function based on said image data and a Vollath'"'"'s F4 function, which is an auto-correlation function, based on said image data;

providing the feature set including the plurality of contrast features as input to a machine learning algorithm that is trained to associate feature sets including pluralities of contrast features with respective position difference values; and

using the machine learning algorithm to classify said feature set including the plurality of contrast features into a single position difference value corresponding to said difference between the sample position and the in-focus position;

wherein

said feature set further includes a sub-set of content features, said sub-set of content features including overall content features and segmental content features, and

said segmental content features are selected from a group including a mean intensity for the foreground segments, a mean intensity for the background segments, a variance for the foreground segments, a variance for the background segments and an area function expressing an area distribution between different segments.

2. The method according to claim 1, wherein said plurality of contrast features are determined for a plurality of color layers.

3. The method according to claim 1, wherein said plurality of contrast features are determined for a sub-set of said image data.

4. The method according to claim 1, wherein said feature set further includes a sub-set of directional features.

5. The method according to claim 4, wherein a number of features of said sub-set of directional features are determined by

identifying differences between a plurality of color layers, and

determining at least one feature value based upon said differences.

6. The method according to claim 4, wherein a number of features of said sub-set of directional features are determined by

identifying a number of different objects having different heights in said sample,

determining differences for said different objects, and

determining at least one feature value based upon said differences for said different objects.

7. The method according to claim 4, wherein said sub-set of directional features includes a quotient between different Laplace-based focus measures according to the formula

8. The method according to claim 4, wherein said sub-set of directional features are determined for a sub-set of said image data.

9. The method according to claim 1, wherein said overall content features are selected from a group including a mean intensity for the whole image data and a variance for the whole image data.

10. The method according to claim 1, wherein said overall content features are determined for a plurality of color layers.

11. The method according to claim 1, wherein said foreground segments and said background segments are determined for a plurality of color layers.

12. The method according to claim 1, further comprising:

determining a number of objects in an object class among the at least one of said object classes.

13. The method according to claim 1, further comprising:

determining a number of pixels belonging to objects of an object class for at least one of said object classes.

14. The method according to claim 1, wherein said feature set is classified into the single position difference value by a support vector machine.

15. A non-transitory computer-readable medium storing a computer program comprising software instructions arranged to perform the method according to claim 1 when downloaded and run on an apparatus.

16. The method of claim 1, further comprising:

adjusting said sample position of the sample based on the single position difference value.

17. The method of claim 1, further comprising:

altering a distance between said sample and an optical system based on the single position difference value.

18. A vision inspection system comprising:

a slide holder adapted to hold at least one slide including a sample;

an image capturing device configured to capture image data depicting said sample when in a sample position, wherein said image capturing device includes an optical system and an image sensor;

a steering motor system configured to alter a distance between said sample and said optical system; and

a processor connected to said image capturing device and said steering motor system;

wherein said processor, in association with a memory, is configured to determine a difference between said sample position and an in-focus position based on a single image by

receiving said image data depicting said sample from said image capturing device,

determining foreground segments and background segments in said image data,

segmenting said foreground segments into segments of object classes, at least one of said object classes selected from a group including a red blood cell (RBC) object class and a white blood cell (WBC) object class,

extracting a feature set from said image data, said feature set including a plurality of contrast features, the plurality of contrast features being generated using at least a wavelet function based on said image data and a Vollath'"'"'s F4 function, which is an auto-correlation function, based on said image data,

providing the feature set including the plurality of contrast features as input to a machine learning algorithm that is trained to associate features sets including pluralities of contrast features with respective position difference values, and

using the machine learning algorithm to classify said feature set including the plurality of contrast features into a single position difference value corresponding to said difference between the sample position and the in-focus position;

wherein said feature set further includes a sub-set of content features, said sub-set of content features including overall content features and segmental content features; and

wherein said segmental content features are selected from a group including a mean intensity for the foreground segments, a mean intensity for the background segments, a variance for the foreground segments, a variance for the background segments and an area function expressing an area distribution between different segments.

19. The vision inspection system according to claim 18, wherein said feature set further includes a sub-set of directional features.

20. The vision inspection system according to claim 19, wherein a number of features of said sub-set of directional features are determined by

identifying differences between a plurality of color layers, and

determining at least one feature value based upon said differences.

21. The vision inspection system according to claim 19, wherein said sub-set of directional features includes a quotient between different Laplace-based focus measures according to the formula

22. The vision inspection system according to claim 18, wherein said feature set is determined for a plurality of color layers.

23. The vision inspection system according to claim 18, wherein said feature set is determined for a sub-set of said image data.

24. The vision inspection system according to claim 18, wherein said feature set is classified into the single position difference value by a support vector machine.

25. The vision inspection system of claim 18, wherein the steering motor system is further configured to alter the distance between said sample and said optical system based on the single position difference value.

26. A control device comprising:

a receiver configured to receive image data depicting a sample when in a sample position,

a processor, in association with a memory, configured to determine a difference between said sample position and an in-focus position based on a single image by

determining foreground segments and background segments in said image data,

segmenting said foreground segments into segments of object classes, at least one of said object classes selected from a group including a red blood cell (RBC) object class and a white blood cell (WBC) object class,

extracting a feature set from said image data, said feature set including a plurality of contrast feature values, the plurality of contrast feature values including at least a wavelet function based on said image data and a Vollath'"'"'s F4 function, which is an auto-correlation function, based on said image data,

providing the feature set including the plurality of contrast feature values as input to a machine learning algorithm that is trained to associate feature sets including pluralities of contrast feature values with respective position difference values, and

using the machine learning algorithm to classify said feature set including the plurality of contrast feature values into a single position difference value corresponding to said difference between the sample position and the in-focus position; and

a transmitter configured to transmit said difference;

wherein

said feature set further includes a sub-set of content features, said sub-set of content features including overall content features and segmental content features, and

said segmental content features are selected from a group including a mean intensity for the foreground segments, a mean intensity for the background segments, a variance for the foreground segments, a variance for the background segments and an area function expressing an area distribution between different segments.

27. The control device of claim 26, wherein the transmitter is further configured to transmit said difference to adjust the sample position of said sample based on the single position difference value.